|kennyschiff ||08-01-2005 01:53 PM |
Reactivations.. Is this common
We have 31 BES users out there, and over the course of the last few weeks several users have stopped receiving email and wireless sync. During these outages users were able to receive and send PIN messages, as well as surf the web. This was across multiple carriers (Nextel, Cingular, and Verizon). The only thing that fixed these was doing a reactivation on the device.
Is this type of outage common? Are there preventative measures that can be taken that would allow for a more consistent service?
Finally, how reliable would BES 4.x be considered as server platform? Our datacenter servces out less than 60 overall users.
Any feedback would be appreciated.
We currently have 35 users and started using BES this spring. We reactivate a unit maybe every week or 2, which seems kind of high-maintenance to me.
It's also had a seriously detrimental effect on reliability of the formerly stable server I installed it on. We have to restart the server 2 or 3 times a week. Bumping the BES services helps, but only a reboot restores normal operation. And then, not for long. If I had it to do over...and I may if I don't find a way to make BES play nice...I'd put it on either a virtual server or a dedicated box. All in all, I'm not real impressed with BES stability at this point.
|kennyschiff ||08-01-2005 03:54 PM |
Originally Posted by JRV
We reactivate a unit maybe every week or 2, which seems kind of high-maintenance to me...
All in all, I'm not real impressed with BES stability at this point.
I'm very concerned about this. Are others also experiencing the same unstability? Can anyone speak to "best practices" to make this more rock solid?
i haven't had any issues at all. i'm not sure on your server resources, but with those kind of user loads, you should not be experiencing these issues, at all. perhaps you're using MSDE and running out of concurrent database connections?
|stonent ||08-01-2005 04:53 PM |
I'm rather interested in the hardware you guys are using to run your BES. I'd like to think that everyone has some nice rack mounted hardware. But I know that isn't always the case.
Our server is a Dell PowerEdge 4600, 2 x 3GHz Xeon w/ HT enabled, 4GB RAM, WS2003SP1, 15K RPM Ultra 320 HDD. RAID 1 on the volume where Windows & BES is running. Not rack-mounted, stonent, but a pretty nice server that should be woefully underutilized given its light load...BES, DC, and light file sharing! And sure enough, it works great...at least for the day after a reboot.
Jibi, we are using MSDE. But...we shouldn't be running out of anything with just 35 users! If we are, RIM shouldn't support MSDE at all.
I've only recently observed that we have the same out of memory error with MDS that others have seen and that RIM has apparently not fixed with SP2, which we've also just installed. Just today I set up a Scheduled Task to bump MDS every day at 3AM. Maybe that will help.
Also, this is an Exchange environment. BES for ABM may be different.
|greywolfe ||08-01-2005 07:10 PM |
just curiously, what BES service pack/hotfix are you on?
greywolfe, if your question is directed to me, we're at BES 4.0 SP2 for Exchange. Just installed the SP yesterday (31 Jul 2005). I am hopeful that SP2 will make some difference.
|greywolfe ||08-01-2005 07:24 PM |
Were you getting these issues before you applied the service pack?
JRV, I suppose it wouldn't hurt to try SQL 2000, if you happen to have a license to run it (or a remote SQL environment already up and running that you can dip onto). With the hardware, you are very much underutilizing it as a BES (depending on your user base, considering the DC aspect). There are stories of people who moved to SQL 2000 from MSDE 2000 and experienced beaucoup increases in performance.
|greywolfe ||08-01-2005 07:58 PM |
Even going to msde2000a might be of benefit. with that hardware setup you shouldnt be getting memory errors or performance drops. I've had environments running 200-300 users on msde without a hitch.
Greywolfe: Problems started as soon as we installed BES (then, 4.0 HF2) to the present. Got worse as we added users. And we already have SQL 2000a.
Jibi: They have a SQL2000 server that only runs an accounting package. Nowhere near as robust a server as the one BES is on, but it's not close to being stressed, either. Plenty left over for BES. But they've been very clear that they don't want any other software touching their accounting server. I've presented that approach several times, including when we were planning the BES install, and been shot down every time. As a result, they have MSDE instances all over the place on other servers for things like Ultrasound, BES, Backup Exec, etc. (With no SMS, it takes forever to update every time MS releases a SQL HF or SP! Looking forward to having WSUS.) I think if I proposed using the SQL server at this moment, given the problems we've had on their file server, they'd fire me!
|greywolfe ||08-01-2005 10:08 PM |
If you've had this problem from the start then I'd say its somehow related to the system setup, software or hardware. A 35 user BES takes up relatively few resources, unless your email load is exceedingly heavy.
Well...I'm open to suggestions. Not much installed on the machine and, as I say, it ran smoothly for a year...right up to the day we installed BES, and as we gradually added users, it gradually got worse. So my feeling is that the problem is indeed software, and the problem software is BES. I'll have a pretty good idea by the end of the week if SP2 or my scheduled MDS restarts helped anything. Next stage of the project will take us to 100 users, but with the server in its current state I'd be real concerned about doing that.
|greywolfe ||08-02-2005 07:43 PM |
I have to say that sounds like a bad memory leak somewhere. Youve obviously updated to the more recent jvm?
Yes, it does sound like a memory leak. I've run PerfMon logs from time to time without coming up with any smoking guns. There was one counter drop that coincided with a couple of BES activations that looked interesting, but I haven't been able to repeat it, so I've dismissed that.
We're at the Java version installed with SP2; 1.4.2.08. But since, anecotally, MDS's memory leak isn't solved regardless of JRE version, I stopped there. Does anything else in BES besides MDS use JRE that would benefit from an upgrade?
That said...2.5 days after installing BES 4 SP2, the server seems stable. That's a recent record. Cautiously optimistic.
JRV, JRE is only really used by MDS's browser push service, if I remember correctly (maybe something else; someone mentioned it here or someone from RIM emailed me about it).
Just curious, since you said this server was also a DC, whats the physical memory load like on this server? I can only imagine that LSASS is sitting at around 500-700MB or more. Add in the BES services to that and you have a constant 1.2GB physical memory charge. Is there another server with less than half the robustness of this one that you can move users to?
|greywolfe ||08-02-2005 09:53 PM |
You installed the BES on a DC? Things like that make baby bes admin cry :(
Originally Posted by greywolfe
Things like that make baby bes admin cry :(
We're not using MDS push. We are, however, getting MDS OutOfMemory errors periodically, and often around the time the server gets flaky. In which case, I guess, it's a leak that happens whether you're using push or not.
LSASS is currently 33MB, 2.7GB of physical memory is currently available, peak commit charge about 1 GB.
Other server choices are limited and there is NO room in the server closet for another server. This client is an SMB about 120 seats in 2 companies that share a system--compromise on system design comes with the territory. One constraint on another member server is about to go away, so it could be a candidate, if necessary. Their only other server I'd even consider for this is their accounting SQL server, but they're adamant that I not use that for BES, or for anything else.
The really encouraging news is that this server remains stable this morning...none of the symptoms we've seen over and over since April. I am starting to wonder if something in BES 4 SP2, or WS2003 SP1, or scheduled MDS restarts might have finally helped us out. I'm crossing my fingers.
Powered by vBulletin® Version 3.6.12
Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.