# Transcript: 754694580
# URL: https://vimeo.com/754694580
# Duration: 4856s (80.9 min)

[0:07] Okay, someone volunteer a question, please.
[0:11] We had had obviously some technical
[0:13] issues with mails and stuff, so people didn't get the invitation. So it will be a small
[0:19] group today.
[0:20] So let's kick it off.
[0:26] If no one else is gonna ask a question, I can ask something
[0:29] which is intriguing In to
[0:32] the March update for SD five five,
[0:35] there was a feature which says account login performance improvements for instance, cluster deployments.
[0:40] But there's a new configuration option called Directory Structure Service Enabled.
[0:47] So my question is, does that automatically
[0:49] create the subscription folders
[0:53] when you create the account?
[0:55] Or does it still require the account to be logged in and then they get created? Which so which folder which parameter was that exactly? Sorry. I pasted into the chat. It was quite a mouthful.
[1:08] Could you go for me a little bit? I don't know if it is the same for everyone else.
[1:14] Can you put the name of the of the variable in the chat for me?
[1:19] Yeah. I have.
[1:21] Oh, okay.
[1:24] Of course, I need to open the chat.
[1:27] Sorry
[1:30] about that. It's early in the morning.
[1:44] Yes and no. So, you so, so the way it works is that
[1:49] when you do something on the admin UI, the admin UI will almost always try to do stuff for you, but not subscriptions.
[1:56] And we don't want to create
[1:58] subscriptions
[1:59] when you do stuff on the admin UI because people make mistakes.
[2:04] Right?
[2:06] So if you do a type, we will create a folder which doesn't make much sense.
[2:10] But before
[2:11] the change, what was happening is that every single time when you log in, even if you log in a 100 times today,
[2:18] we will check if every single folder is created for you. So if you have
[2:23] a thousand subscriptions,
[2:25] we'll check a thousand folders
[2:27] every single time you log in.
[2:30] K.
[2:32] With the change,
[2:33] we are now caching the information about that. So if
[2:37] we had checked that the folders are there, we
[2:41] assume that they are there. So during the login, we don't spend the time to recheck everything if we think you are okay.
[2:47] It still can backfire a little bit. Obviously, someone is going on the OS level and deleting folders,
[2:53] and that's why it's a parameter.
[2:55] But if you don't touch your OS level and you know your admins are not deleting folders because they have nothing better to do,
[3:02] then enabling the feature
[3:04] allows
[3:05] for,
[3:06] the cluster to actually
[3:08] cache the information in the memory, which means a full restart will basically wipe out all that stuff.
[3:15] So it starts over.
[3:17] But if you have
[3:19] a huge account that logs in thousands of times to put files in
[3:24] and therefore the structure never changes,
[3:27] that saves all that time because it does a single check at the beginning, marks it as okay, and then goes on.
[3:35] Okay.
[3:37] Hello?
[3:38] That's especially important on HTTP.
[3:41] So if you're on SSH,
[3:43] the only folder we really, really care about is the one you're in at the moment where you will put the file. Right? So we technically can even check for the folder existence on CD if you move out
[3:55] or on a list. Right?
[3:57] HTTP is not like that. HTTP is, loading the whole structure on on login.
[4:04] So
[4:05] make sense a little bit?
[4:07] Yeah. Yeah. It does make sense. I was from the wording, I wasn't sure whether it would create all the box subscription folders up front, which is what would have been.
[4:16] Part of remember that TM and admin have a little bit different permissions occasionally?
[4:24] So allowing the admin to create folders can create troubles, especially on rest of the environments.
[4:29] Okay. So we generally will leave TM to do the work, which doesn't mean that the home folders will not get created occasionally.
[4:36] There is code over there, but it fails about 90% of the cases in my experience,
[4:41] fails gracefully.
[4:43] And but the change itself is not in the decision if it is done in our team doing it. Nothing changed there. The change really is if you log in, check if we already had checked your folders
[4:56] soon ish if you're in in memory already.
[5:00] Yes. The first login will take the same amount of time as before, but subsequent ones. Yeah. And it saves a lot of time.
[5:07] The reason this was introduced, by the way, was a customer. They have a single account feeding about half of their files into the system, and they log in separately for each file.
[5:18] And one of the accounts that they are using for that has about 7,000 folders, subfolders.
[5:24] Okay. About
[5:26] 200 of them were subscriptions.
[5:29] So every single time they're logged in, we had to jot off 200 folders,
[5:34] which, as you can imagine, was adding a lot of stress to the file system.
[5:38] So that's what this is all about.
[5:41] Okay. Thanks for explaining.
[5:44] Makes sense? Yep.
[5:46] Okay. So, Annie, on on this thing, if if your shop just mainly
[5:50] has scheduled
[5:52] SSH transfers,
[5:54] this,
[5:55] this parameter really wouldn't come into play.
[5:58] It will, actually.
[6:00] Because remember that there are two places where we check if the folder exists. One of them is on login.
[6:06] The other one is during a pull.
[6:09] You know, both actions that actually
[6:11] can get the file to.
[6:14] It's less likely to be needed, Joe. But if you have accounts
[6:19] with a lot of subscriptions,
[6:21] I'll still enable it.
[6:23] Yeah. Because I I look at my subscription my my subscriptions or accounts that have, you know, at least in our our shop that are bigger,
[6:30] they they never log in because it's all it's all done by the scheduler
[6:34] and all by the TM.
[6:36] So that's why I was thinking. It doesn't really
[6:39] right?
[6:40] Yeah. The way it's named, and I haven't checked if it's doing that also on,
[6:47] let me think.
[6:49] During a pull, the only folder we will check will be the one you're pulling into anyway.
[6:54] So, one folder more or less is not a problem.
[6:57] Still,
[6:59] if you are not touching your folders from the OS level, if you don't have CleanupScript deleting folders and so on, I'll still enable it just for the few cases where it will matter. Sure.
[7:13] I mean, it's
[7:16] it's not about your configuration in next year much more more about
[7:20] who is getting to your storage area?
[7:23] Because we have customers that have other fingers on the storage area, everyone knows me and know what I'll say next.
[7:29] Right?
[7:30] If you have your fingers in my storage area, you're doing it wrong.
[7:35] Just let the ST do whatever ST does best.
[7:39] But if you don't have anyone like that, enabling or disabling it will not change anything for you except that it might speed up things a little bit.
[7:47] Okay. Thanks.
[7:50] And if you look at the name of the value, it's directory search service enabled. It has nothing to do with BingQ on login. The login itself is the most
[7:59] used event for that, but there is a secondary check during the pool as well. Remember, if you set up a pull and the folder doesn't exist, we'll create it. Right?
[8:08] This is the same check, basically.
[8:14] Worth enabling it, or at least test with it, see how it behaves.
[8:18] If it doesn't trigger, then it doesn't make any difference, but you still have at least a few people up load
[8:25] sorry, logging in. Right?
[8:27] Or is everything automated these days?
[8:31] Yeah. For our shop, we have especially SSH. I mean, every some users coming in to do HTTP uploads or downloads, but, you know, it doesn't hit my my deployment.
[8:42] Okay. Well, for those HTP guys, this will be useful. So enable it. K.
[8:49] It's it's harmless. It's one of those things that if you are not in the situation where we're checking, nothing changes for you. So
[8:59] and in case you wonder how you know if it's breaking or anything, if people start complaining they cannot see their folders, that's what, just disable it again and check with support what's going on. But I haven't heard any troubles with it.
[9:14] Okay.
[9:17] What else do we have for today?
[9:21] Good morning, Annie. This is Lisa.
[9:24] Hey, Lisa.
[9:25] And Hi. Lisa,
[9:27] just give me a second. Nay, you're next.
[9:32] Okay. At least we can go first.
[9:35] Okay. With me. So
[9:38] question, so go ahead.
[9:40] Then I was gonna ask Okay. So, Lisa, go ask your question first.
[9:44] Go ahead, Lisa.
[9:47] Okay. So we're having an issue. We are on the June patch.
[9:52] Our
[9:53] environment is clustered,
[9:55] and it's on a Windows platform.
[9:58] When we restart services
[10:01] or recycle services,
[10:04] and the primary comes fully up,
[10:08] and then the secondary the secondary
[10:11] both servers thinks it think they are primary.
[10:14] And we have to do multiple restarts
[10:17] and, even kill a Java process in order to resolve that issue.
[10:24] Yeah.
[10:26] You're on standard cluster. Right?
[10:31] Sorry. Lisa? This is standard cluster. Correct?
[10:36] So
[10:37] how it works
[10:39] is that
[10:41] when a server comes up live,
[10:44] it connects to all its partners
[10:47] and asks,
[10:49] Is one of your primary at the moment?
[10:52] And if you don't have a response,
[10:54] or
[10:55] if
[10:57] a response times out
[10:59] or some or someone responds and says, Not yet,
[11:02] then it self declares as primary.
[11:06] So if your secondary self declares,
[11:10] there are
[11:11] two possibilities for a problem.
[11:14] One of them is you have an old Java process that never get cleaned out on the restart, which I suspect you already checked for.
[11:24] It can be a networking problem,
[11:26] But basically, what's going on is that the communication between the secondary and the primary
[11:31] is
[11:32] getting
[11:35] not correct for some reason or another. And I will blame Windows for that.
[11:41] Okay.
[11:43] Did it start after the update, or was it happening before?
[11:50] I believe
[11:52] Anai, you can answer that. You or Marita can answer that. You guys do the patching.
[11:57] Yes. It was after the five,
[12:00] the the 5.5 upgrade, but it's on the after the patch, that last patch, the June 30 patch.
[12:07] The June 30. Okay.
[12:09] I haven't heard about issues about that,
[12:12] but it's possible that something in the patch which was trying to optimize something actually kicked it off.
[12:19] One thing to try
[12:21] will be,
[12:22] so
[12:23] when do you start the secondary? Do you wait for the primary to say become primary?
[12:30] Okay. So,
[12:33] we did a, an OS
[12:36] patch yesterday.
[12:37] Uh-huh. And we had, this is on a
[12:40] a stage environment testing. But
[12:44] we patched the secondary first,
[12:47] and then, you know, after the secondary was patched and it came back online.
[12:52] Now the primary
[12:53] has not been patched yet. So the primary
[12:56] once the secondary comes back online,
[12:58] it sees the primary
[13:00] as offline. The secondary sees the primary as offline,
[13:03] and then the primary sees the secondary as,
[13:07] you know, also, online.
[13:10] Is that right? Yeah. You know what I will say. Right? This has nothing to with ST.
[13:15] During the OS patching,
[13:18] either your channel between them got closed or something changed on the kernel level. I don't know what to tell you. If it happens after the OS level patch,
[13:28] my good feeling is the two OSs are not communicating well to the networking at the moment, or at least not fast enough immediately after the patch.
[13:36] So
[13:38] try to do so if it is an s so try to do the following. Shut down the whole system, both secondary and primary, for secondary then primary.
[13:46] Bring up the primary only.
[13:49] Look at the server log and wait for a log
[13:52] entry that says become primary is called or something along these lines.
[13:57] It's a one word, become primary.
[14:00] And this is the function that that says, I am the I'm the server that is in the cluster, no one else is primary, and I'm the first one, so I'll call it.
[14:09] Once you see that, start the secondary and see if what you're seeing will persist.
[14:15] Okay.
[14:17] So when you're doing patching and we're on a cluster, do you recommend patching the secondary first or the primary first?
[14:27] Does it matter?
[14:28] It doesn't matter, really. They should be able so the only thing to think about is if your secondary ST is the only several life even for a second, it will become a primary,
[14:41] and it will take over the queue and the event table and everything else.
[14:46] So if you have the choice, especially if you can take the outage, what I would recommend is to shut the whole system down,
[14:54] touch the primary first,
[14:56] get it applied on its own so it's the only server in business,
[15:01] and then touch the secondary.
[15:04] Okay. That way, you don't have the flip flop of the,
[15:09] primary, secondary, and you don't end up with a couple of transfers going with the secondary as a primary because then their tracking will be on the other tracking table.
[15:19] But in this case, what you are the way you explained how it happened,
[15:24] I think something on the networking level was still connecting when ST came
[15:29] up live. So when the secondary tried to reach to the primary,
[15:34] it literally couldn't get to it,
[15:37] or it timed out.
[15:40] Okay. So
[15:41] all aspects can be a pain in the behind. And I know we're recording, but this is an official term, technical term or should be.
[15:50] Especially on Windows, where Windows is trying to help you by closing security stuff occasionally.
[15:58] And,
[15:59] you know, I I will not tell you how many times I've been on the phone on Windows installation. It turns out that patch enabled
[16:06] the firewall, the Windows firewall again
[16:09] even though it had been disabled.
[16:12] So
[16:14] if it still persists
[16:16] and it doesn't work, get support to look into it. But at this point,
[16:20] this is not the release the ST release itself is my gut feeling, quite honestly.
[16:27] Something on the OS patch is influencing us.
[16:31] And it might be temporary during the startup, the Forrester startup,
[16:34] or
[16:35] who knows?
[16:39] Excess?
[16:41] I know it's not what you want to hear, but
[16:45] it's
[16:47] very
[16:48] mhmm. Go ahead.
[16:50] I think, yeah, that makes sense, Annie. Because of the two different OS patch levels, I can see the contention
[16:57] and where it would each would think it's primary.
[17:01] So
[17:02] okay. Yeah. It's it's also possible that they were still communicate. Remember that we just have a timeout in there. So it's possible that the two OS patch servers on the two OS patches were actually trying to talk to each other and still communicating and resolving issues between them. And in the meantime, we declared it because our time out came.
[17:21] Right? Right. It's possible that the new OS patch got some security a little higher on the on the secondary,
[17:28] so it needs some time to start communicating with the older box. You know?
[17:32] And we cannot write all that because we literally just come up and we try. And if the connection fails, we assume that there is no server on the other side.
[17:41] Also, look at the server lock. It should tell you if it didn't connect with the timeout or if it errored out.
[17:50] Right?
[17:50] Because this secretary
[17:53] can't reach to the primary to check if the primary is there. It doesn't matter how you call them, but any
[17:59] server coming live on a standard cluster, first job out of the door, connect to everyone else on the cluster, and check if someone is a primer.
[18:09] That's his job.
[18:11] Right?
[18:12] So Right.
[18:14] There should be a lock telling you if it was a timeout or an error.
[18:18] And it it can be something
[18:20] stupid at the end. Did it connect at the end, or is it still not working?
[18:25] Oh, it's connecting. To work.
[18:27] Yeah.
[18:29] Yeah. It it's probably about timing. Also,
[18:32] are your servers starting on startup of Windows, or are they starting immediately after the startup?
[18:39] Did you leave the services on automatic,
[18:42] or do you have them
[18:44] on manual and have a special service that calls start off?
[18:52] Nay, do you know? Lisa
[18:55] will probably know better.
[18:57] Yeah. So we do we
[18:59] do have them on automatic,
[19:01] but we're having issues with that also.
[19:04] Some,
[19:05] start automatically,
[19:07] and some services do not. So okay. And that's your problem.
[19:13] Basically,
[19:14] on Windows,
[19:15] Windows is too greedy, so it will start try to start any service it can as fast as possible.
[19:21] The problem on the SQL, you need order.
[19:25] So,
[19:26] you need the database first, right? And if the database is still down, you have a bigger problem. Obviously, this is not what's happening here. But it's possible that the transaction manager, when it started the first time,
[19:38] its
[19:39] networking was not properly connected, or the networking of the whole box was not yet connected. Some of the networking services of the windows were not connected yet when TLIF
[19:48] has started
[19:49] reaching for its partner.
[19:52] So, what we recommend on Windows is to not have the services on automatic,
[19:57] but instead to build a separate service which calls stack all instead,
[20:02] And that runs at the end of the process that
[20:05] and that gives enough time because by the time the database picks up and all starts and by the time you get to the TM to start, which is usually the last one to start,
[20:14] if you look at the sequence of start all.
[20:17] By
[20:19] the time the TM starts, everything else on the box had fired up.
[20:23] Right.
[20:24] So and that's why
[20:26] I'm saying I think it's the OS patch because especially because you're on automatic, chances are the TM just started way too fast and you still had networking services
[20:36] booting and rebooting, and that's what was going on over there.
[20:40] And again, look at the server lock of the windows, it will tell you who started when, but you also know that the TM will declare itself started about two minutes before being ready.
[20:49] Maybe it just got ready a little faster this time around. Who knows?
[20:54] So
[20:55] it's an order tank.
[20:56] And when that happens,
[20:58] a stop all start all will cure it almost always as long as you wait it out,
[21:03] and you kill everything. And I heard that you mentioned that you had the Java process as well you had to kill at one point.
[21:10] Yes. That's the same problem. Basically, something started so fast that it disconnected on its on its own, so you it basically died on you.
[21:19] So
[21:21] that's the problem with Windows and automatic
[21:23] services.
[21:27] Hi, Annie. This is Edie. So when you do a stop all, should that
[21:32] command
[21:33] also stop that Java
[21:35] that Java task in Windows?
[21:37] Oh, yeah. It should. If it doesn't, you had had a problem during the startup, and something got disconnected.
[21:45] Okay.
[21:46] So stop all should stop absolutely everything except in the old servers. It was not stopping the
[21:54] schedulers.
[21:57] That was the only process to say,
[21:59] okay,
[22:01] my brain is not giving me the correct word,
[22:04] whatever effect we're using for schedules and scripting the UNIX-one.
[22:08] That was the only one that was not stopping and had to be killed manually.
[22:12] But in the 5.5 releases,
[22:14] a stop all on any platform whatsoever
[22:18] should stop every single process running from out from inside of the ST folders,
[22:24] which includes
[22:26] all the JAVAs. You know, you have a lot of them running.
[22:29] So if it doesn't,
[22:31] and that that most likely means that someone is starting out the folder,
[22:36] which might not be a functional problem,
[22:39] but might be something to watch out during restarts.
[22:45] And again,
[22:46] as I mentioned, if you're going to put a service on automatic, instead of putting the individual services on automatic, I will just have a separate service that just does that all.
[22:56] It buys you enough time usually.
[23:03] Okay.
[23:05] Makes sense?
[23:08] And I know Windows is just different from everything else we are running on. But Windows is also the only one that you literally can't order the services around.
[23:19] So
[23:20] it's trying to help you. That's the problem with Windows.
[23:24] It assumes that everything in the automatic list is
[23:28] essentially
[23:30] can start at the same time, let's say, like that.
[23:34] And it's not always true.
[23:38] I have another question that we run into with patching and our cluster. So on 05/21,
[23:46] our VIP
[23:48] VIP is configured as active standby.
[23:52] Mhmm. So when the primary cluster
[23:56] goes offline,
[23:57] it starts moving the traffic to the secondary, which is now been promoted to the primary.
[24:03] In five two one,
[24:05] the VIP looked for
[24:07] a port.
[24:10] But now
[24:12] now in five five, it doesn't
[24:15] that port is no longer
[24:17] a valid
[24:19] trigger for our bit to start
[24:23] passing the traffic to the active primary.
[24:27] Is there a way for that to happen? Because, like, right now, we're we're not a truly
[24:33] clustered environment.
[24:35] It was looking for it was looking for 4455,
[24:38] isn't wasn't it? Correct.
[24:41] Yeah. Unfortunately,
[24:43] when they reversed
[24:45] the TM communication.
[24:48] So back in the days, it was edges or demos talking to a transaction manager or 54455,
[24:55] so you could monitor 4455.
[24:57] Right? Right. When they reversed it that now the TM is opening, this listener disappeared.
[25:03] And without that listener, you don't have an active port.
[25:06] So
[25:07] there isn't an easy way to figure out if you switched over, unfortunately.
[25:17] Let me think a little bit on that. I'll get back to you on that one. Because the answer is nope. No easy way.
[25:25] And,
[25:27] but okay. So if you're doing that way with the okay. Let can I ask a stupid question?
[25:33] I mean, what you're describing is active passive, not active active?
[25:38] Yeah. It's active well,
[25:41] our network folks call it active standby.
[25:44] Well, yeah. But from ST perspective,
[25:48] this is the configuration you'll do for an activepassive cluster, not for an activeactive cluster.
[25:54] Okay.
[25:56] So
[25:58] I'm not sure why they are doing it that way. I mean, your your cluster is set to active, right?
[26:06] Yes. Alicia, will you confirm that?
[26:12] What was the question again, Annie?
[26:14] Your cluster
[26:15] type is set to active. Correct?
[26:18] That's correct.
[26:20] Yeah. So in this case, you might want to talk to your VIP people because,
[26:27] I think I know what they're doing. You don't have a load balancer up there, do you? It's just the VIP, or do you have a load balancer?
[26:35] It is
[26:37] not a load balancer. It, it Yeah. That explains it. Yeah. They got cheap.
[26:42] That's what's going on.
[26:44] So, usually, what you'll do is instead of having just a zip like the way you have it, you'll have a load balancer, and then you do a round robin or weighted or something like that, thus allowing inbound traffic to go to both boxes all the time.
[26:59] Because at the moment, what you what you have is your
[27:03] primary
[27:04] is receiving all of the inbounds,
[27:08] while your secondary
[27:10] is
[27:15] only helping with
[27:19] outbounds and the stuff happening from the server. And you hate me for saying that, but you should actually,
[27:25] change that a little bit. You want your secondary to be your main server for the inbound section.
[27:32] The primary is handling folder monitors, schedulers,
[27:36] the event queue, the tracking table, all of the heavy lifting behind the scenes.
[27:40] In addition, you are also giving it all the inbounds.
[27:45] See where I'm going with that? If they will need to do the VIP the way they are doing it, it actually should be sitting on the secondary, not on the primary.
[27:56] Okay. Still thought?
[27:57] I,
[27:58] that is a really different thinking than what I understood. So could you help me out a little bit more on that?
[28:06] Yeah. So, okay, because you're in an active cluster,
[28:10] both so let's forget about your VIB situation for a second. Let's just talk about secure transport on its own, and then we'll scale out of it. Okay? So in the SQL world,
[28:20] when you're in active active, the two nodes are
[28:24] working in conjunction.
[28:27] Both of them can accept inbound transfers.
[28:30] Both of them can handle outbound
[28:32] and server jobs.
[28:35] The big difference is that the primary is also responsible for a lot of the maintainers. So it does the event queue.
[28:43] It does all event distribution. So if there is a pull to be done, the primary is the one pulling from the database. It looks around and says, I have
[28:52] now I'll run it, or it sends it to its secondary.
[28:56] It does the folder monitoring, all the scheduling, you know, everything that is server side
[29:02] originates always on the primary.
[29:05] Then the primary can send some of those jobs out for the secondary for processing and listening for staff.
[29:11] So, the primary TM handles most of the jobs and a lot of the work. So, if you look at distribution of file transfers,
[29:19] you will usually see the primary handling anywhere between
[29:23] 6080%
[29:24] of everything
[29:26] with the secondary assisting with the rest,
[29:29] unless you start getting overloaded, in which case you'll see fifty fifty. In this case, what we're talking about is just your outbounds. Your inbounds are stuck because of the bid.
[29:39] So far with me?
[29:41] Yep.
[29:42] Okay.
[29:43] So
[29:44] let's think about where your so this was for the outbound. So at this point, you have a primary server twice as busy as the secondary, not just before we transfer, but with all of the additional staff.
[29:55] The scheduling,
[29:56] the scheduler process, the folder monitor process, the maintenance job, everything is working on the second time. It's the one every time when there is a transfer going through the cluster, the secondary just tells the primary, I got the file,
[30:10] here are the details, and the primary writes to the tracking table. So, is the primary always doing all the writing to the tracking table,
[30:17] right? So, most of the jobs are sitting on the prime.
[30:21] With your VIP on top of everything else,
[30:25] All of your inbounds are going now to the primary.
[30:30] So let me ask a question. Are they going to the primary server, or either are they going to the to a primary edge?
[30:38] So this is our internal
[30:41] system. Yep. They do not have edges. No. Edges. And that's what I was thinking is happening. Because now what is happening is so the primary server has let's call it 60% of the jobs already
[30:53] from the server jobs.
[30:55] Usually higher, 60 to 70.
[30:57] The secondary will have at this point 30 to 40. Now you put all inbounds
[31:02] on the primary. So in addition to all the additional stuff, now you get also all the inbounds.
[31:08] When the file arrives on the SH daemon on server one
[31:12] or HTTP daemon or something,
[31:15] the only TM that can talk to it is the local one.
[31:20] So, it will get processed and received on the primary always.
[31:24] Now, the incoming go to the secondary, and some of them will go because incoming can't use a server job. Remember that.
[31:31] But the fact is that the busier server is also your inbound server.
[31:38] Okay.
[31:39] So,
[31:40] what we usually do in this case, what when people need to do it the way you do it with the DIPs, what I usually recommend
[31:46] is to get the inbounds to go to the slower server, which base to the secondary,
[31:54] that you don't load your primary with both the server jobs and the client jobs.
[32:03] And instead,
[32:05] you utilize your secondary beta.
[32:10] Okay. So that's really different
[32:13] I know. Than my than what
[32:18] That that's because you are think thinking of the secondary on ST as a backup for
[32:22] ST, and it will be two if you're in active passive mode. Because in the passive mode,
[32:28] the secondary is just sitting and waiting to take over if something goes wrong.
[32:32] But in the active active, you actually literally have two full blown servers working with you.
[32:39] Makes
[32:41] sense? A little bit? Little bit. Yeah. Yeah. So usually
[32:45] usually what we do is to have a load balancer instead of your just a bit. And then in the load balancer, it's not around Robin, but it is weighted toward the second time.
[32:57] So that you send, for example, 30% of the inbound into the primary, 70% to the secondary, something like that.
[33:03] And I cut off someone.
[33:07] Oh, yeah. So, in your recommendation to send it all all the inbound traffic to the secondary,
[33:14] is that an active passive connection
[33:17] or still active active?
[33:19] From the ST perspective, you're in an activeactive.
[33:23] On the VIP level, it will be a,
[33:26] failover like they do at the moment except that it will be a reverse failover.
[33:32] They'll mark up the second yacht secondary as their primary.
[33:37] Oh, okay.
[33:39] And the reason why you need to do this gymnastics is because you don't have a load balancer, because you're on a VIP, and because the VIP can only be in a failover mode,
[33:47] so they can only go to one server. And I'm saying if they have to choose during normal operations, I'll choose the one that is less busy at the moment, which always will be the second.
[34:02] And I'm not saying you need to do it today, just food for thought, because what you have at the moment,
[34:08] you have self limited how many files can go through your system or how many connections we can have, right?
[34:15] So if you start having performance issues, that's that's your first place where you can tweak things so that your performance get a little better and get more files through.
[34:25] If the current system works, don't don't, you know, not not broken. Don't touch it.
[34:31] But if you start having performance issues, that's where you start.
[34:37] Well, it is kind of broken for us in the fact that it,
[34:42] if the primary does go down,
[34:45] there,
[34:46] there's no there's no continuity.
[34:49] So
[34:50] Yeah.
[34:51] Down,
[34:52] the secondary gets promoted, but there's no traffic going to it. And so now we have an outage. So I know. And that has to do with how we do the Vips, and I will think about what we can do.
[35:04] Because without a load balancer up there, most of the ways we do it now are impossible to be done because there is no port for it.
[35:13] Right. Okay.
[35:15] Okay. One thing, by the way, which is a little stupid, but I've seen people doing it,
[35:21] is to run a cron job or something like that. And then now we're on Windows, so it will be obviously
[35:27] a scheduler
[35:29] that, checks who is the primary at the moment.
[35:33] And if it is not,
[35:35] the primary shuts down one of the protocols, the one that you can monitor for that protocol instead of the four four five five.
[35:43] Okay.
[35:44] So, for example,
[35:46] you you
[35:47] can even use one of the protocols you are not using, PECID, for example.
[35:51] You know, keep it up on one of the ports on PECID or, you know, the multi configuration.
[35:57] And then you have a process running every five minutes. You'll still have an outage. It will not be automatic, but it it can check locally if the TM is running or is the TM who is primary at the moment.
[36:11] And if it finds
[36:12] that, yep, I'm not the primary,
[36:16] if it had switched,
[36:18] technically speaking,
[36:19] it cannot go back to the back to the primary without the manual sync anyway.
[36:24] But it can switch off.
[36:26] It basically can issue stop visit or whatever protocol you're using. And then if that's what they're monitoring, it will work.
[36:33] Okay. So that I know that you
[36:37] wouldn't consider that a best practice, but it sounds like that is a
[36:42] the the
[36:43] streamline
[36:44] Mhmm. Or some fastest
[36:46] thing that we can put in place to
[36:51] to to get us a a truly clustered environment
[36:56] from our perspective and then, work with our,
[36:59] network team to get that done on the VIP side. Yeah. That's the best thing you can do because you don't have a load balancer. If you had the load balancer,
[37:08] you have a lot of options. You can run a file. You can do a ping. There is a lot of things load balancers can do. The VIP switch is literally,
[37:17] you know, check if the port is up. That's all it's checking. Right? Right. So just put it on a different port.
[37:24] Right.
[37:26] And it Well And because and we know that would work for you because you used to monitor four four five five, which was not the port they're forwarding anyway because they don't forward ports anyway.
[37:36] So,
[37:37] yes, it will require running a separate service or process on the Windows boxes, but,
[37:44] you know, it's
[37:46] should be easy enough.
[37:48] Yes.
[37:50] Yes. Actually, this this bubbled up. I think when we first
[37:55] implemented
[37:57] five five,
[38:00] our network team
[38:04] was was
[38:08] monitoring port four four three.
[38:11] So it didn't look like there was, there was an issue until
[38:15] the just the TM went down. Four four three was still up, and it did not it did not move it over. So that's kinda Yeah. That's why, yeah, that's why you need this additional service running that basically checks and then shuts down the protocol.
[38:30] Right. So that's yeah. And and, of course, that means yeah. Sorry.
[38:35] That means that you need to remember when you switch it back on to start the protocol again.
[38:40] That will tell the VIP that you're back live.
[38:45] But, you know, let's just
[38:47] write it down and, you know
[38:50] Yes. Not the best way, but it will work, especially because you have empty protocols that you can use for that. Right. Yes. Okay. Good. I mean and and you can always use SSH or and HTP as well. You know? I just don't like to stop them and start them when I don't need to because they take a couple of minutes to start occasionally. Right? Right.
[39:09] So
[39:10] but starting and stopping the PECID daemon that you never use or the AS2 daemon for that matters. You know, one of the two that you don't use for anything,
[39:18] if you don't use them,
[39:20] shouldn't be that much of a problem because, yes, it will start slowly, but no one is looking for it anyway. All you're looking is if the port is listening.
[39:28] So who cares? Right.
[39:31] So, yeah. So, I mean,
[39:35] this is the basic idea.
[39:37] Can you make it nicer?
[39:39] Probably. I mean, talk to your OSRMs.
[39:41] Talk to your networking people. But something like that should work. Yeah. Okay. Great. Thank you.
[39:48] As for how you determine who is the primary,
[39:51] there is a Sentinel Sentinel
[39:54] whatever file of okay. My brain is not working today.
[39:58] There is a file in, one of the temp folders in ST. There is an API call.
[40:04] There is a lot of ways to check if this server is live at the moment. You can even run a file to it. You know? Log if you can log in over SSH or HTP or one of the protocols, if you can log in on locally on the box, the TM is up and running,
[40:19] and
[40:20] that means you can use that one. So
[40:24] how you want to implement that script that checks and shuts down and shuts down the port that the other guys are listening to, that's a long conversation, and you find the best way. It will not be immediate, so you'll still take,
[40:40] downtime,
[40:41] you know, if it fails. So if you're checking every five minutes and it fails immediately after the previous check, for the next five minutes, there is no one. But in five minutes, it will be automatic.
[40:51] K. So you can do it every two minutes or something. You know? It's about tolerances, but it's better than what what you have now where you need a physical person to figure out that something went wrong.
[41:03] Yes.
[41:04] So, I mean, five minutes down is bad, but five minutes down is better than three hours down without anyone realizing.
[41:11] Right.
[41:13] Exactly.
[41:14] So think about something like that.
[41:18] The other option and,
[41:23] other options, other options, not with the VIP. With the VIP, you don't have that many other options.
[41:28] So talk also to them if they can read a log file. So talk also to the VIP people. If they can monitor a port, is there other things they can monitor?
[41:37] Can they monitor a log file?
[41:40] Remember that there is a very specific
[41:43] entry in the log file called become primary,
[41:47] which if someone can monitor for,
[41:50] you can send it to the file system and the database at the same time so they can monitor there. Okay? And they call an API, you know?
[41:59] Start from there as well. Because if they can actually use one of the mechanisms you will be using in order to shut down the port itself,
[42:07] you don't need to inject the secondary service.
[42:11] Yes.
[42:12] So I just don't know what they have set up in their environment because
[42:16] in my experience,
[42:18] VIP switches are very, very stupid. They literally can't check a port, and that's about it. Yep. But who knows what they have built in your environment? So check with them.
[42:31] K?
[42:32] Good.
[42:33] Thank you. Mhmm.
[42:36] Okay.
[42:38] What else do we have?
[42:46] Joe, you have anything interesting in your world?
[42:51] Things are pretty,
[42:53] pretty slow right now.
[42:55] So, unfortunately, nothing
[42:57] nothing major to report or talk about.
[43:03] Okay. Question?
[43:06] Okay.
[43:07] Good.
[43:08] Maorita?
[43:09] Okay. I can't pronounce your name. I apologize.
[43:12] You you pronounced it correct.
[43:15] If there's no other question, I just wanna ask about the,
[43:20] the transcode. Sorry. Let me just go back to that. The the op when setting up a transfer site, the transcode any line terminator
[43:29] in asking mode option
[43:31] where we can check and uncheck that one.
[43:33] Other than
[43:34] other than when we check and uncheck, is there any additional
[43:39] condition
[43:40] where
[43:42] where
[43:44] this this option
[43:46] will kick in or not, meaning to say,
[43:50] if we if we leave it unchecked,
[43:53] based on the documentation, it says it's supposed to add the CR
[43:56] on the end of line, right, for this one, for FTP
[44:01] ASCII.
[44:03] Yep.
[44:05] When when we were testing on this one, I see I see that happen if we're transferring
[44:12] outside secure transport to an, to a Unix or to an IBM,
[44:17] or a mainframe. Right? But if we're transferring between
[44:21] between the two secure transport we have, I don't see it happening.
[44:27] No, it doesn't because we are Windows to Windows.
[44:31] Basically,
[44:33] when you run a transfer in ASCII mode
[44:36] Mhmm.
[44:37] The receiving server
[44:41] is responsible
[44:42] for,
[44:43] putting the end of lines and stuff like that.
[44:47] So what protocol are we talking about? Open server. I have a server here somewhere.
[44:54] So which protocol you were playing with?
[44:57] This is for an FTP,
[44:59] FTP and then ASCII.
[45:02] Okay.
[45:02] So let me Oh, go ahead. See what happened.
[45:07] Okay.
[45:09] Sorry about that.
[45:11] You know, sessions expire.
[45:22] Going to an account.
[45:25] I don't have here is my account.
[45:28] See, I'm very
[45:30] predictable. Everyone knows how my accounts are called.
[45:33] And we're doing an FTP one. Right? Mhmm.
[45:36] I'm not going to save it. It's just so that we can all look at what we're talking about because everyone is visual. Right?
[45:47] And you mean this one. Right? Yes. ASCII. Okay.
[45:51] So what do you we sent it to ASCII.
[45:54] All it does is basically to tell the other sort of so, okay, let let's step back a second. And I I know I'm doing that a lot, but it's easier for me to explain when I start from the beginning.
[46:06] So the difference between an ASCII and binary mode is that when you connect to the other server, you tell them that this is a text file or not. Right? So when you tell them that it's an ASCII file or text file or whatever you want to call them, you are telling them,
[46:21] This will you
[46:24] need to strip whatever end linings I have, regardless of what type they are, and put your own.
[46:34] While in binary,
[46:36] it means transfer the file as is. So what happens in ASCII mode? If you're going from Windows to Unix,
[46:43] on Windows,
[46:45] you have CRLFS
[46:48] and one.
[46:49] Mhmm. Unix has only CR. So what happens is when when CRLF
[46:55] so when SD sends the file,
[46:57] and here I'm not exactly sure if we're going to strip the serial, we're not stripping the serial, but this is an end of line, so the other server knows to replace it with TR immediately.
[47:07] So if you're in ASCII mode and you're moving from Windows
[47:10] to Unix, the
[47:13] CRLF
[47:14] will get replaced by a CR.
[47:17] If you're starting from Unix
[47:19] going to Windows in ASCII mode, the opposite happens.
[47:24] So far with me?
[47:26] And and that only happens and that only happens with between secure transport.
[47:33] But when we are No. No. It always happens. It has nothing to do with ST. That's how the protocol is supposed to work.
[47:41] Okay. So this is this is what we noticed on our testing.
[47:44] From our, so our external,
[47:47] transport is running in Unix. Right? And then our internal is in Windows.
[47:52] So we transfer the file we transfer the file
[47:56] we transfer the file from from a Unix to
[48:01] from the Unix to to the
[48:04] Windows.
[48:05] It did it did what you're saying. It it con it it it
[48:10] if we have the trans whether we have the transcode check or uncheck that that option,
[48:17] it disregarded
[48:18] that. So it it changes it changes the, end of line into CRLF
[48:23] because the destination is Okay. Okay. Question.
[48:26] So when you say you haven't set it up, if it is auto detect it or auto detect it, If you set to binary and it still does that, open it to the desktop.
[48:36] No. Binary, it's doing the binary, it keeps the end of line if we set it to binary. But if we set it to
[48:44] so that transcode,
[48:46] it it it it doesn't seem to do what what was,
[48:49] what we thought it does where it's
[48:52] and so
[48:53] it still converts the the the end of line to Windows.
[48:58] When we transfer possible.
[49:00] Yeah. But when we transfer
[49:03] from the Windows Secure Transport to any of our other server internally,
[49:08] then that transcode
[49:10] seems to affect the file. It it's it's
[49:13] if we leave it unchecked, it's adding the CR.
[49:16] Let's say a window I'm sending a, let's say I'm sending a
[49:21] Windows
[49:22] file to a to a Linux server. Right? So it's supposed to it's supposed to convert that into,
[49:30] LF only,
[49:31] for the Linux server. So it's it's gonna convert.
[49:34] I send CR only, but okay.
[49:37] I'm sorry?
[49:39] CR only. It's not LF only, but okay.
[49:42] Okay. So it, it sends it to to the server.
[49:46] But Uh-huh. It becomes Windows again because,
[49:49] if we leave that transcode unchecked,
[49:52] when it gets transferred,
[49:54] even though it's ASCII,
[49:55] when it gets transferred to the Linux server, it will Okay. It will gonna have the CRLF.
[50:01] But if we Okay. If we checked it, it will gonna deliver it as whatever is post the end of line. Okay. What is the name of the file? What is the extension of the file? A TXT.
[50:14] Okay.
[50:16] So
[50:17] if a text file in AutoDetect
[50:21] is not behaving the same way as in ASCII,
[50:25] get support to look into your server. Something is not set correctly somewhere on the OS level.
[50:31] It's supposed to. The idea of the AutoDetect is it will grab the MIME types out from your OS level,
[50:39] and a text file will always be a text file.
[50:43] I see. Yeah. So
[50:45] so so the the short answer is if you go in binary mode, nothing should change on the file. If you go in ASCII mode, the file arriving
[50:55] on the other side
[50:57] should have the end linings of the receiving server.
[51:01] If you go in AutoDetect,
[51:03] it depends on the extension.
[51:06] Mhmm.
[51:08] So if it is dot TXT,
[51:10] it should be a text file. If it was dot doc, we're not going to touch that one.
[51:17] That's why I asked you what the extension is.
[51:19] Yeah. We've, yeah, we've been testing, that TXT and that TXT file. Yeah. So if if you can reliably
[51:27] reproduce
[51:28] it, open a ticket to support and explain to them exactly what you have where,
[51:34] because something somewhere is misconnected.
[51:36] We had historically
[51:38] troubles with the FTP not picking up ASCII properly
[51:42] occasionally,
[51:43] maybe one of those books
[51:45] showed up again.
[51:48] So that's
[51:50] the best I can say. But the theory is the rule
[51:54] basically is that in ASCII mode, as long as it's just to ASCII. Oh, the other thing to check,
[52:00] open the tracking table.
[52:02] You know? And I don't have anything in this one. Oh, I might have. Hold on. Let me see if I have something.
[52:16] Yeah, I don't have anything. But if you open the tracking table on the outbound,
[52:22] it will tell if the file went in ASCII or binary because this is your first point of contact. Yes. If it says
[52:29] it's binary when it's a bot. So if it was auto detect,
[52:32] then if you go to the tracking table, it will tell you what it picked actually.
[52:36] So if it says it went into ASCII mode, but a Unix server received a
[52:41] Windows file, that's a bug somewhere. Open a ticket, get support to look into it. Okay. Yeah. It, yeah, it's it's
[52:49] yeah, it it it went out in ASCII. So so you're saying so you're saying that transcode that transcode,
[52:57] if we set it to ASCII, it's
[53:01] It will always treat every single file going through the system, through this transfer site as a text file.
[53:09] And it will
[53:10] initiate the transfer in ASCII mode, so the receiving server is supposed to strip and replace the endpoints.
[53:17] So that transcode doesn't do any supposed to be not doing anything anymore?
[53:22] That transcode check on the transfer side?
[53:25] Oh, it does.
[53:26] What it it controls
[53:28] if we open in Ask Your by the remote.
[53:33] If all your files are text files
[53:36] and you stay in Auto Detect, you are right. It doesn't do anything interesting because it's the same as if it was asking.
[53:44] But if your file is called ANI without an extension,
[53:48] this file will by default be a binary. If you want to force it to be an ASCII because you know they're all text files, you need to switch to ASCII
[53:55] hard coded into the site. Mhmm.
[53:59] That that's what this is all about. It's literally
[54:02] how you open the connection. It doesn't change the file. It doesn't do anything from the file. What it does is to set
[54:08] the connection in ASCII or binary mode for the for the transmission,
[54:13] which tells the receiving server how to handle the file and what to do when it sees end appliance.
[54:20] That's it.
[54:25] Which means if you have a transfer site that you always know they're just text files and you never ever receive anything but but text files,
[54:34] I will set it to ASCII just
[54:36] to be on the safe side in case something happens.
[54:39] Yeah. We we we do, sorry if I keep going back to this one. We do No. You're okay. We do set it to ASCII. We we did set it to ASCII. The, and then when we checked the outbound lot, it is set to ASCII.
[54:52] But what my concern is is that option, if you, if you go back to a transfer site Mhmm. There is that option, to check and uncheck the transcode.
[55:02] It says transcode
[55:05] transcode any line terminator in ASCII mode.
[55:10] Yeah.
[55:12] So don't read too much into it. I think you're just reading a little bit too much into it.
[55:18] Because,
[55:19] it it
[55:21] because when we, so if you do FTP and then ask you,
[55:26] and then right below the preferred passive mode, there's that,
[55:30] transcode any line
[55:32] that That's
[55:33] a okay.
[55:34] That's a different thing. Yeah. That's the one I was actually asking, on if you have any additional condition,
[55:43] that it it it triggers or not. Because when we whether we checked it or not between our two secure transport, it doesn't seem to it doesn't seem to affect.
[55:53] But if we're transferring outside secure transport, that that's when we when we notice
[55:59] that it it it it makes a difference. Because after we migrated to s t five five,
[56:05] we started,
[56:06] for the for the FTP ASCII, we initially have set up. They started seeing extra lines in between because CR CR was added.
[56:17] Then
[56:18] and then let's say it's going to
[56:24] Linux. Right? So their CR and then the whatever the original extent I mean, the the destination
[56:30] end of line character is there, but an additional CR is added.
[56:35] So when we checked it, it resolves the issue.
[56:39] But, but it's not consistent.
[56:43] Yeah.
[56:44] So,
[56:45] okay. This one is a little weird because what it does is basically get ST
[56:51] to do something on them before going to the remote server, and it's not consistent because
[56:58] your receiving server is not consistent.
[57:01] So some of the receiving servers require the transcoding to happen before they receive the chain, before they receive the connection.
[57:09] Some of them can work with the with the raw file itself.
[57:16] So for some of them, it will need to be enabled for some disabled. So for this transcoding,
[57:21] it is when ST
[57:24] changes
[57:25] the file before sending it out. So if you don't check it and you're coming from Windows,
[57:31] still the end of lines are going as they are.
[57:34] Okay?
[57:36] With the transcoding,
[57:37] it's changing it changing it for for the end server so it's easier for them to realize what's going on. You will almost need to play
[57:47] with every server you ever talk to with this checkbox and see when it will need to be enabled or disabled. It depends on the other server,
[57:54] for the most part. So yes, for a lot of servers, it will not make a difference if you check it or not.
[58:00] I see. And we and
[58:03] that and like you said, it's depending on the destination server.
[58:06] Yes.
[58:07] And sorry that, I was going on for the other
[58:10] field, but it actually makes sense because it's still so think of it this way. You have two partners during the transfer, right? The client, which is ST, sending,
[58:20] and the receiver where the file goes or comes from if you're doing a pull, right?
[58:26] So both of them need to talk, and sometimes the receiver actually needs a little bit of help. That's why transcoding is in play.
[58:33] But when you need to enable it will depend on the other server.
[58:38] Okay.
[58:39] That's why we have a checkbox. We didn't use to, if you remember,
[58:43] because we never needed to kill, but some servers are just weird.
[58:46] Officially,
[58:47] if they're RFC compliant, you shouldn't we shouldn't need to do anything. We sent the Windows file onto your Unix system. It should recognize it and do it.
[58:57] Who knows?
[58:58] And people build their servers clearly.
[59:01] Yeah. It it's just that it, it was a surprise because with,
[59:04] with the five four, it was never an issue. But then when we, moved to,
[59:10] five five, it became it became,
[59:12] it became
[59:13] Okay. That we have to,
[59:15] to to verify. So
[59:17] if you have a specific use case that is reproducible
[59:22] that used to work on 5.4 but works differently on 5.5,
[59:26] get support on it.
[59:30] Even if you can solve it on your own by playing with the value,
[59:35] we cannot test every single server in the world, right?
[59:38] So support having information of this happening might help our in this spot what they did. It's possible that we change the behavior somewhere.
[59:46] We upgrade remember that all of this stuff is based on external libraries,
[59:51] and we keep upgrading them,
[59:53] all right? So if an external library changes how it does things, sometimes ST needs to do something to
[60:00] to compensate it. Maybe we've missed something in there.
[60:05] Right? We wouldn't know until someone reports it to us.
[60:10] So if you see a change in behavior,
[60:13] do talk to Axway.
[60:15] That's important, as in open a ticket.
[60:18] Okay, thank you. Even if it is to say, okay, guys, so this is how it works, work, it used to work on five point four, this is how it works now on five point five, I make it work this way,
[60:29] but this is a change in behavior. Is this expected?
[60:33] You know?
[60:35] Just like an advisory.
[60:36] It's like partnership.
[60:38] Right?
[60:39] Half of the stuff I know is because customers
[60:42] screw it up and I need to fix it. That's how it works. Right?
[60:46] The other half is because I screw up and I need to fix it.
[60:50] Okay. But more often.
[60:53] So, Annie, this is Edie. So you're you're saying that it depends on the receiving
[60:58] server.
[61:00] And Yes. So
[61:02] can you is there, like, a rule of thumb
[61:04] to, like, a Windows
[61:07] or a Unix?
[61:08] No. It's not about the OS. It's about the the software vendor that built the server
[61:15] and how close they were following the RFCs
[61:19] and the optional sections in that.
[61:23] Okay.
[61:24] It's a so all of the FTP and FH servers on the market are supposed to be RFC compliant, those things,
[61:31] right,
[61:32] except for the big guys sometimes.
[61:35] Occasionally,
[61:36] a company decides to be let's call it creative.
[61:39] I would use other words, but we are recording and that's a professional environment,
[61:43] I'm not going to use other words.
[61:46] So let's call them creative.
[61:48] Those creative servers,
[61:50] you need to play around them.
[61:54] In my experience, the transcoding
[61:56] stuff usually is required for old servers.
[62:00] But this is just my
[62:02] very
[62:06] subjective
[62:07] idea simply because that's the kind of stuff I usually get put into.
[62:11] So, you know, servers that were built twenty years ago and never updated on stuff like that.
[62:17] Yeah. So like what, Marita said, so
[62:20] on 05/04,
[62:22] we
[62:23] we pretty much
[62:25] had everything that we knew that was a text,
[62:29] especially everything going to our mainframe.
[62:32] We forced it ASCII,
[62:33] and then we put that transcode because we do have
[62:37] Unix servers in our environment that we pull from.
[62:41] So we just put the tran set the transcode.
[62:45] Now moving to five five, it that changed everything. Some files were were fine. Other files were
[62:53] added an extra line, which
[62:57] which
[62:59] when Scope of everything. Yes. Caused errors.
[63:03] And we couldn't we can't we still can't really find a pattern.
[63:08] And, I mean, it's
[63:10] I think we upgraded,
[63:14] like, months ago, and I think we've finally
[63:18] iron,
[63:19] sorted all that out. Our concern is that when we patch,
[63:24] you know, one of those libraries are gonna change,
[63:27] and then,
[63:29] you know, our FTP
[63:31] transfers with text files are gonna be thrown off again.
[63:34] So
[63:35] So, okay, so 5.4 to 5.5 was a huge jump,
[63:40] as in probably seven different versions of this library went through.
[63:44] This
[63:46] I'm not going to promise you it will not happen again,
[63:49] but it's a lot less likely to happen again.
[63:53] Okay.
[63:54] It's
[63:56] that's the risk in not to so part of the reason why we tell, I will always tell people, try to plan for updating every six months
[64:05] is
[64:06] that the chances of libraries jumping that
[64:10] much every
[64:11] six months is a lot lower than anything else.
[64:15] Part of the and and you said it's mainframes. It's the old servers. That's what's going on. Technically speaking,
[64:22] every time when we update security and we cannot not update security,
[64:26] there will be at least one old server that will not like us anymore.
[64:30] So you will need to do some gymnastics to get it to work.
[64:33] Right.
[64:34] And this is the risk we're all running, but, you know, there are two choices. We all stay in the ninety nineties and somehow run through the servers and hope for the best on security,
[64:46] or you make
[64:47] your server as secure and as modern as humanly possible
[64:52] and then deal with the fallback end. If it means that you need to tell a partner that, you know, it's a twenty first century, come join us.
[65:02] It's and I know how hard it is. And,
[65:05] honestly,
[65:06] the mainframes will always be your biggest problem. I mean, some of those machines are older than me at this point,
[65:12] but they work.
[65:14] Yeah. I think we I think, actually, the mainframe,
[65:18] files
[65:19] actually
[65:20] got resolved the quickest.
[65:22] Mhmm.
[65:23] And we kinda saw the pattern there. It's,
[65:26] it's the
[65:28] other environments
[65:29] that some But that's that's Some didn't need need any change, and then others,
[65:37] it took some work to dial in.
[65:39] Yeah. Because you don't know what exactly they're running on other side with the mainframes. One, you got the pattern, all mainframes are fixable, you know, you just do the same change.
[65:49] But when you go against other systems,
[65:52] they will be if you pull up the list of what they are running on the other side, you'll see the patterns there as well.
[66:00] Usually.
[66:01] It's
[66:02] that's that's the, you know, it's a communication.
[66:06] And FTP,
[66:07] and I'm sorry to say it, but there is a reason why everyone's height is is running from FTP, not just the double channels and everything.
[66:15] The protocol is just so
[66:18] widely defined that this kind of stuff happens all the time.
[66:25] You know? Yes.
[66:28] Someone touches something, and everyone starts crumbling around because
[66:32] I don't know if you had ever read the RCS for FTP because I did that a few years ago.
[66:38] There is so much stuff in there that, you know, if you do that and do that and do that and it's a Tuesday.
[66:44] Well, they don't say that, but it almost sounds like that.
[66:48] You know? It's just a huge protocol,
[66:51] and it's also a legacy protocol at this point. Right.
[66:56] And it's we're trying to and trying and part of the
[67:01] so when SSH was built, the SFTP protocol was built,
[67:05] it was built with security
[67:07] as a
[67:08] first thought, you know, you cannot have non secure as FTP.
[67:13] Right. FTP security
[67:16] was a secondary thought to the protocol. Its first job was move the files.
[67:22] But in order to secure it and make it more reliable and make it guaranteed, because, you know, these double channels are pain to track,
[67:31] they keep adding all kinds of jingles on it and all kinds of weirdness, and that's what's going on with it. And that's why if you have a problem, it will be on the FTP level. Unless we screw up really badly on the SH side, 99%
[67:45] of the cases, SH is clear.
[67:48] FTP on the other second?
[67:51] I mean,
[67:55] and I know you need to use it because of how internal environments work, right, and so on, but it is the reality of it.
[68:07] But yes, if you have a repeatable scenario that shows the difference between five point four and five point five,
[68:13] either open a ticket in support, or if you just don't want to do that, send me some information
[68:19] with exact details so I can write up something because you're not going to be the only one in this role.
[68:26] Okay,
[68:27] thank you.
[68:28] So, and again,
[68:30] we test as much as we can, but there is just too many of them out there.
[68:38] Thank you. Morita, that okay for you? Yes. I I
[68:42] know that you would love an answer that says, yep, just check that in this specific case, and you're all done.
[68:49] Yeah.
[68:50] Our
[68:51] job doesn't work like that, does it?
[68:57] But and,
[68:59] you know, now that you know which broke on the previous update, you know also which ones to keep an eye on immediately after the next update.
[69:07] That is that.
[69:09] Thank you.
[69:11] Okay.
[69:13] Okay.
[69:15] Do we have anything else, ladies and gentlemen?
[69:18] I have a question. This is Nye.
[69:21] Yep.
[69:23] Photo monitors.
[69:25] Now if
[69:26] we're
[69:27] we have quite a few photo monitors transfers.
[69:31] And
[69:33] if we need to stop using photo monitors, what other protocols would you recommend us to use?
[69:39] And is there a specific a number of protocols
[69:44] that I mean, photo
[69:45] monitors that we should be using?
[69:48] Okay.
[69:49] You can have as many folder monitors as you want. There are two things to be very careful about. Number one,
[69:57] if you have
[69:59] a single folder monitor
[70:01] looking for
[70:03] so,
[70:04] not okay, if you have multiple
[70:06] folder monitors
[70:08] inside of the same folder looking for different patterns, for example,
[70:13] it's a big issue on the on the file system level. So, in this case, consolidate into a single one,
[70:22] and then use advanced routing to distribute.
[70:25] But
[70:27] I am yet to find
[70:29] a number of older monitors that break the new servers. Back in the five to one days,
[70:34] I could have told you exact numbers because it was the old implementation.
[70:38] We rewrote the folder monitor.
[70:41] I have customers running about 5,000 of them without trouble.
[70:46] That's good to know.
[70:48] Oh, also,
[70:49] one thing that we did add.
[70:52] So the new folder monitor
[70:55] has,
[70:56] and it's not new anymore, but you you know, it was sometime in the 05/02/1953
[71:01] releases. But,
[71:03] if
[71:04] you
[71:05] know that files are arriving only at the specific times, keep in mind that you don't need to leave the folder monitor as a real monitor. You can also use it as a scheduled monitor,
[71:15] which means that if you know that files will only be arriving between four and five, there is no point in turning 20 fourseven.
[71:21] So you can use a scheduler.
[71:25] So folder monitors can run it, so for you know for any other transfer site it's always schedule based.
[71:30] With folder monitors you have two separate modes.
[71:33] One of them is scheduled, the other one is unscheduled. The unscheduled is basically the constant monitoring every five seconds.
[71:42] And it is used for cases where files can arrive at any time, or when there is a lot of files coming all the time. But if you are using it in the way you use an SH site, for example, you want to check every hour if there is a file, you can do that on folder monitor as well. Those folder monitors,
[71:59] which are on the schedule,
[72:01] are working differently from the ones on non on the standard ones,
[72:06] which means that when we're talking about how many you can have, those scheduled ones don't count.
[72:12] Oh, okay. That's good to know. Because the slowness
[72:16] and the problem with a lot of folder monitors doesn't come from the process of the folder monitoring. It comes from this constant monitoring every five seconds.
[72:26] Okay? So when you move the scheduler into
[72:30] a scheduled
[72:31] one,
[72:33] running every hour, or if you move it to every two minutes, you don't do it much of a change. Right? In this case, I would say leave it on to normal. But if you're checking every hour or
[72:43] only twice every day or something, this is like an s h side for us almost in terms of performance.
[72:51] And, again, I'm still to break an s t server or five five server with too many folder monitors, monitors, and I've tried very hard.
[72:59] Okay. So unless you're thinking about having 10,000 of them tomorrow or and the biggest use case you need to make sure never happens is
[73:10] multiple folder monitors in the same folder.
[73:13] That's your biggest problem.
[73:15] And it's not on the SD level, it's on the file system level.
[73:19] Okay. That's good to know. Because you have too many. In this case, again, just consolidate and the advanced routing can distribute the files. So if you are more looking for files starting with any underscores and files starting with underscore
[73:31] and so on, just make a single folder monitor looking for all the files,
[73:37] and then use advanced routing to distribute them.
[73:40] Okay. That's the basic rule.
[73:43] And it's the same with pools and everything, you know. But,
[73:46] if you're thinking about adding 10 or 20 of them, don't worry. Now,
[73:52] if there is someone that is producing 10,000 files at the same time
[73:57] and you need to pull them with a folder monitor, that will slow down.
[74:03] Okay. So We all know that.
[74:05] We do notice that sometimes that, you know,
[74:08] secure transfer,
[74:09] you know, move the file into the subscription folder and transfer them to the destination locations,
[74:16] but sometimes it still holds the keeps the file in the subscription folder.
[74:22] Yeah. You have a there is a checkbox for that on the subscription level
[74:27] when you set up the when you
[74:31] so
[74:33] look at the subscription level. There is a checkbox at the very bottom what will happen with the file on success.
[74:44] Subscription
[74:45] is you know, there's the options to, you know, delete it,
[74:49] but sometimes it doesn't always delete the file, and it lives there.
[74:54] Which means that it errored out somewhere.
[74:57] So so when one of those happens, what you want to do is to see to go to a tracking table and the server log and see what happened. Chances are
[75:06] something errored out in the last stages. Even though the file was transferred, it was considered an error, not a success.
[75:15] So we will manually move those files depending on how whether it's fully tracked. Yes. So so so the the subscription basically gets a single,
[75:26] status back from the route in advanced routing. The route tells it success or not.
[75:31] If it's a success, then it turns the post transmission on success. If it's an error, it turns the one on error. So, you need to figure out what happened. If you have it enabled
[75:42] on the subscription level, you know, at the very bottom I was sharing because I told them.
[75:47] So,
[75:51] share.
[75:52] You see my screen, right? Yes.
[75:55] Okay.
[75:56] Patricia,
[75:59] subscription.
[76:01] That's a wrong
[76:04] one because I'm in the wrong place. Hold on. My bad.
[76:12] Doesn't matter. But over here at the bottom?
[76:16] Nope, that's, hold
[76:18] on. I keep picking up the wrong one.
[76:21] That's the correct one. This was shared folder. Shared folders on here. So here at the bottom, the post routing,
[76:27] that's what we're talking about.
[76:29] If this is set,
[76:31] the file will disappear for the subscription file. Yeah. We always have it set to delete it on success. But if I was using the basic applications, even then, I would still, you know, set it to that. But, I mean, not then. You know, we do see file that gets stuck and then get So if okay. So if they get stuck,
[76:48] either it came out as an error, so it basically didn't go to that mode,
[76:53] or because you're on window doing those,
[76:56] and that might be what's going on,
[76:59] that we could not delete the file because of the Windows didn't allow us to.
[77:05] So, when that happens,
[77:08] try to investigate to see what happens.
[77:11] Try to see if ST didn't try to delete or if the deletion was blocked internally. If it was blocked internally, you go to file your file system people and tell them to get their antivirus or whatever they're running off your storage.
[77:22] Remember at the beginning of the call when I mentioned something about people and fingers on my data?
[77:29] On remember how Windows works. That's important, actually. On Unix, even if they have a scanner on top of their seat, they don't log to files.
[77:37] But on Windows, if they are scanning, for example, the area where the files are arriving, even if they think they are not doing a problem, doing anything,
[77:46] chances are occasionally they might be locking the file we're trying to work with.
[77:51] That's why we don't allow scanners.
[77:53] So when it happens next time,
[77:56] find out if it is ST not trying to delete or ST unable to delete. The log file will tell you which choice you are in. If it turns out to be a testability one try, that means the route came out as a failure, not as a success. So look at your route
[78:12] and see what might be happening in there and look through the log file for that.
[78:17] If it looks like ST says that everything is
[78:21] okay and the deletion succeeded because if the OS tells us it deleted, it's okay, but the file might still be there. If this happens, you need the OS logs of the Windows and your OS admins to look through them at the approximate time to see what happened.
[78:37] Okay,
[78:38] thank you for that. So connect the dots. You know, find out who is not doing the deletion. As long as this is enabled, and make sure it's not just mistakenly forgotten, but as long as this is enabled and the route reports as a success,
[78:53] this delete will trigger
[78:55] or should trigger. If it doesn't, talk to support. If it does but doesn't delete the file, that's an OS level issue. It's not us.
[79:03] We try.
[79:04] Because all SD can do is to try to delete it. It should delete command to the file system. If the file system doesn't take it or responds, yep, I deleted it while it didn't,
[79:13] guess what?
[79:14] We don't know that. We don't care. We don't check is the file still there or not. We believe the file system.
[79:20] Okay.
[79:21] So
[79:22] we're trusting people occasionally.
[79:28] Makes sense,
[79:29] I hope? It does make sense. Okay.
[79:34] Okay.
[79:35] We're almost on the
[79:37] bottom of the hour. What happened here? We have been talking for hour and a half.
[79:42] So
[79:43] Amazing.
[79:45] Yeah. Yeah. It happens with me all the time. So,
[79:50] last
[79:51] call for questions.
[79:54] There's
[79:55] nothing in the chat. It was just Morita's Well, thing that you had to
[80:00] that's part of it is that four of the the four ladies are actually from the same company.
[80:04] Yeah.
[80:05] They were already going to save the environment for me. So, Joe, any additional questions? Last No.
[80:11] Any thank you so much.
[80:14] We'll, as as Nicole said, we'll probably set up another one in a few weeks because, obviously, something else did not make it through. Darren, you okay?
[80:25] Yeah. I'm thanks, honey.
[80:27] Okay. Ladies?
[80:32] Thank
[80:33] you so much.
[80:34] Yes. Thank you very much.
[80:46] Sorry about that.
[80:47] Okay.
[80:49] Thanks, everyone. Have a wonderful day.
[80:53] Thank you, Annie.