Has anyone had windows server repeatedly blue screen after installing Stash?

[Kevin Wood]

I installed Stash on a Windows Server 2008 R2 according to the installation instructions and the server has been blue screening anywhere from one to five times a day since. When my system administrators looked at the memory dump they said Tomcat was the problem (CRITICAL_OBJECT_TERMINATION). I had multiple Tomcats running but as of the last crash, only the ones for JIRA and Stash were enabled. I have since disabled Stash and have not had a crash in almost 24 hours. I would like to continue using stash but can not unless I find a solution to this. I have tried both Stash 1.1.1 and 1.1.2. I am using 32bit Java 1.6.0_33 and the git linked from the installation instructions, currently Git-1.7.11-preview20120710. I have Stash running as a service as per the atlassian instructions. Thanks to anyone who might be able to help.

[Stefan Saasen · Accepted solution]

We've identified the root cause for this problem and released Stash 1.2.4 with a fix for this.

Please see https://jira.atlassian.com/browse/STASH-2719 for the details.

Stash 1.2.4 can be downloaded from: http://www.atlassian.com/software/stash/download

Cheers,

Stefan

[AjayA]

Hey Guys,

We've managed to reproduce the BSOD locally on a Windows Server 2008 RC2 VM. Implementing the suggestions from the following Microsoft KB Article appears to prevent the BSOD from occurring in our environment:

http://support.microsoft.com/kb/2535804

We've also had positive reports from customers who implemented the fix.

Can you please try the KB and let us know if this fixes the BSOD?

We'd love to help! Please raise a support ticket on https://support.atlassian.com

Cheers,

Ajay.

[Wade Bennett]

Can confirm that this is occurring on a physical machine as well. I've also had Tomcat crash without a full Windows crash.

[Lars Hannelius]

We have now also tested on a physical machine running 2008 R2 with the same result (BSOD). I guess VMware is off the hook.

[AjayA]

Hey Guys,

This is pretty high on our priority list and we're actively investigating this issue; I'll have an update as soon as we find the cause/fix for this issue.

Regards,

Ajay Sridhar.

Atlassian Support.

[Kevin Wood]

I am running FishEye but not Crucible on the same server. I disabled Stash for the time being and have not had a chance to try 1.2.0 yet as Lars-Erik suggests below. I have had the issue on both virtual and non-virtual servers.

[PeterL]

Are you also running Crucible / FishEye on the same server?

[PeterL]

I'm getting the exact same error under essentialy the same conditions.

[Lars Hannelius]

(stash_bsod_minidump.txt) I've experienced a similar BSOD twice on a virtual (VMware) Windows Server 2008 R2 instance running Stash as a service. The first time it happened a few hours after the initial install of Stash 1.1.1 & Git-1.7.11-preview20120620. The second time it happened a few minutes after upgrading to Stash 1.1.2 & Git 1.7.11-preview20120710. In both cases the minidump analysis (attached) indicates that the crash happened in the Tomcat process but it doesn't say where exactly. I'm using a 32-bit JDK 7u5.

The latest upgrade to Stash 1.2.0 appears to have gone ok.

I'm suspecting it could be a VMware problem but right now I don't know what to do about it. I'm just hoping the server won't start crashing in between Stash updates.

[Lars Hannelius]

I have raised the issue with Atlassian support.

One thing that we noticed is that the BSODs seem sensitive to the VMware coresPerCpu setting. Our VM is allocated 8 virtual CPUs and we've tried running with the following configurations:

A. 1 socket with 8 cores per socket
B. 4 sockets with 2 cores per socket
C. 8 sockets with 1 core per socket

We can easily get the machine to crash in configurations A and B by running a script that replays some random clicking activity in Stash. Configuration C appears stable thus far but we're obviously very concerned whether it'll hold up in the long run.

[Mike McGuire]

Was this ever resolved, Lars?

I have the same issue running Stash 3.10.2 on Windows 2008 R2 in VMWare. My minidump results are nearly identical to yours.

Thanks,

Mike

[Stefan Saasen]

Hi @Mike McGuire,

This was fixed in the past but there was a regression that affects your version. Upgrading to at least 3.10.4 should fix this. See https://jira.atlassian.com/browse/BSERV-7620

Cheers,

Stefan