MSN Messenger drops frequently

[rtcanathan]

Our CSR team uses MSN Messenger in different office buildings, and it seems that recently, if the Untangle VM is running, the Messenger program will logout frequently. If I stop the untangle VM, the problem goes away.

I haven't made any changes in Protocol Control for MSN Messenger, other than the originally setting it to "flag" months ago. I have had some other recurring issues that seem to be resolved with a reboot of the Untangle server itself, but I'll address that later in another thread.

Does anyone have any suggestions on where to start looking, or what may be causing this? No settings have been changed, it just seems like to starts imploding after a certain amount of uptime, and I can't really see anything that sticks out as a specific culprit.

[sky-knight]

Other than that is normal behavior for MSN? I use MSN heavily as well, and I gave up on the stock client and started using pidgin because it does a better job of maintaining presence.

If you are logging IM sessions that may slow things down enough to cause a timeout if your box is underpowered.

[rtcanathan]

Originally Posted by sky-knight

Other than that is normal behavior for MSN? I use MSN heavily as well, and I gave up on the stock client and started using pidgin because it does a better job of maintaining presence.

If you are logging IM sessions that may slow things down enough to cause a timeout if your box is underpowered.

I don't think the box is underpowered, everything excedes the min recommendations. When I run a top, all resources look exceptionally well. I have a scheduled restart for noon, and after the VM is back up, I'll run another top, and screen shot it. If just flagging the protocol would affect it like this, I would think my problems would be more widespread. I'm not ruling out a physical problem with the box, it just seems odd that a restart seems to clear things up for a while, and then problems seem to slowly compound until I just turn the VM off until I can restart it again.

I'm only blocking very few things at the moment, for a baseline of what office/user traffic looks like, but I am flagging a moderate amount of things so that it shows up in the logs. I can't imagine that would be enough to cause my problems, it would be a bad thing if it is.

[sky-knight]

"Min Recommendations" from where? System requirements for Untangle are a moving target, and if you're running something that only barely meets the min requirements with more than 5 people behind it that is the issue.

That said, MSN is a world of inconsistency all by itself. The UVM is going to inject some delay in packet processing as a natural side effect of the way it works. If your connection isn't quite quick enough to begin with it's quite possible to have it drop just because of the added UVM latency.

That said, I am honestly grasping at straws.

Try disabling protocol control and see if the issue goes away. It's entirely possible that your box just can't log all of the sessions and keep them live. I doubt you need to disable the entire UVM. You've shown the issue lies in the rack somewhere, now it's time to isolate which module. I'd start with Protocol Control.

[rtcanathan]

Originally Posted by sky-knight

"Min Recommendations" from where? System requirements for Untangle are a moving target, and if you're running something that only barely meets the min requirements with more than 5 people behind it that is the issue.

That said, MSN is a world of inconsistency all by itself. The UVM is going to inject some delay in packet processing as a natural side effect of the way it works. If your connection isn't quite quick enough to begin with it's quite possible to have it drop just because of the added UVM latency.

That said, I am honestly grasping at straws.

Try disabling protocol control and see if the issue goes away. It's entirely possible that your box just can't log all of the sessions and keep them live. I doubt you need to disable the entire UVM. You've shown the issue lies in the rack somewhere, now it's time to isolate which module. I'd start with Protocol Control.

I feel confident in ruling out the bandwidth (we are an ISP and our pipe to the office is good enough that I can say with 99.999% certainty thats not the case).

The box could be, its an older 1U supermicro that we wiped and loaded with untangle, but its fairly robust for this application. mb= PDSMi, 4gb ram, mostly just deadstock from supermicro. I don't have all the specs, but its probably just a p4 server proc. But still should be enough for this application I would think.

We have less than 70 connections coming through it, usually less than 40. I really don't think its the culprit, unless there is a memory leak that shows its face over time, but I have my doubts to that as well. It would be odd that the server would start to implode like that, without showing any significant resource commitments when doing a top. Before I restart, I'll turn the VM back on, screencap top, restart, and screencap another top. There won't be any real resource commitment change (other than the RAM commitment by the VM itself, which support said is normal, obviously).

I kind of figured this issue would be in the protocol control, and can easily unflag alot of things, but I don't have that many enabled. Mostly chat, games, and peer to peer are flagged (and the ptp is also blocked as well). I wouldn't think that would cause these issues, but maybe it will. Its disappointing if it is.

And generally after a reboot, the thing will be ok for a week, maybe two? Then it starts going down again. I had some strange stuff with the web caching, so I finally just turned that off and left it off. I hope I don't have to slowly strip this down so that its only doing the bare minimum stuff.

[sky-knight]

That's an awful long time between failures to do any real testing.

How long as this unit been in service?

[rtcanathan]

I would have to say since March or April, somewhere in there. Seems like the problems are getting more frequent and more service affecting too, and I really only have minimal configs made.

Since my configs are pretty minimal at this point, I'll just wipe the thing and reload it with the latest/greatest release and start from scratch.

[sky-knight]

That isn't a bad idea, some upgrades from 8.1.1 to 9.0 didn't go so smoothly. That said, I'm not sure that is really going to fix anything. However, it surely cannot hurt.

[rtcanathan]

Maybe before I do wipe it, I'll give it a week with alot of the logging and ptp blocking disabled. Then re-evaluate then and see if in fact that is the cause of my problems. It'll drive me up the wall if it changes for the better after a wipe, and I don't know exactly where it was.

[sky-knight]

Welcome to the club! Format C may cure the symptoms but it sure doesn't give you any more information as to what the issue actually was, and you have no control if the issue returns.