Zen Intranet Install fails (on Windows server 2008r2+MS sql)

[Nic Hart]

I have been unable to install Zen Intranet/Lightning on a clean Windows server 2008r2+MS sql Confluence on 5.1.

When adding the sample content the installer consistently hangs during the sample projects install. And Confluence hangs for all users. Only way to restore service is restart Confluece service (and on one occasion needed restart of Windows).

Failed 3 times now....

[Nic Hart · Accepted solution]

I'll close this question since I have characterised and worked around the issue to the extent it is no longer blocking my progress, and hopefully my advice my be of value to others.

The solution I found was to simply keep repeating the cycle of restarting Confluence and retrying the 'Configure' option for Zen Intranet. At each cycle I always found at least one more item on the manifest was installed, and as far as I can tell when eventually the process completed the resulting site was not defective.

I also found that lower powered/smaller Linux instances exhibit the same problem. In every case you can tell that a restart is needed by monitoring the CPU activity, which falls to zero when the install hangs.

And to answer the likely first question from the support team - yes, there was plenty of Java heap and permgen free - though it did seem to slightly improve matters by allocating more than the default. But even an 8CPU/16G RAM Windows instance with half the RAM given explicitly to the JVM still needed three attempts to complete the install.

[Kathy]

Hi Nic,

I would like to pursue this issue to find a better workaround than multiple restarts. If you have the time, I *would* appreciate your AWS server config notes.

I don't believe memory usage is the issue -- the payload is just too small. It was developed on vastly less capable hardware than the 8CPU/16G instance you tested, and at no point did we see resources clobbered on any of our test instances (which are combo of dedicated hardware and AWS vms).

I wonder if this was a problem with database comms (hung instance with quiet CPU), so I want to replicate your problem on the same AWS config, and browse the sys and sql logs.

So, again, if you can find the time, I would appreciate your AWS config details (kathy@stepstonetech.com).

Either way, thanks for the follow-up repoort!

[Darryl Duke]

Hi Laszlo, Thanks for opening a ticket at support@stepstonetech.com. We'll continue the troubleshooting through that channel.

[Laszlo Kremer]

MSSQL Server Developer Edition (64-bit), Windows server 2008r2 standard x64, Confluence 5.1.3

[Kathy]

To our knowledge, this problem is occurring with MS-SQL Express only. If you're seeing this problem on full MS SQL, would you mind specifying the version?

[Laszlo Kremer]

Happens the very same to us as the OP described. Memory seems to be enough, as it uses about 700 MB when the install stops (it stops at every space it want to install).

Bonus: you can't unselect the remaining spaces :)

[Darryl Duke]

Hi Martin,

We're suspicious that MS SQL Express is the culprit, as it's not a database supported by Atlassian. However, on the https://confluence.atlassian.com/display/DOC/Database+Setup+for+SQL+Server page, search for SQLExpress and in the comments you'll find the following note:

For anyone out there using SQL Express, I found this thread on Stack Overflow to be helpful:

http://stackoverflow.com/questions/1862283/help-me-create-a-jtds-connection-string

You need to add ";instance=SQLExpress" to the end.

Perhaps this is related to the the problem you're seeing...

[Martin Schaible]

Hello,

I have exactly the same issue here. I can reproduce the issue on several machines (Windows 2008 R2 with MS-SQL 2012 Express).

To finish the setup, Tomcat needs to restarted at least 4 times and it's really annoying.

I have tons of exceptions in the log, which must have to do with this issue. Please contact me, if you like to have the log.

[Kathy]

Hi Nic,

Sorry you're having these frustrations -- this is the first report we've had of install problems.

If you continue to see these problems, please open a ticket via email to support@stepstonetech.com. I'll want to get some file system information as well as screen captures and logs, to try to understand what your server is doing with the content payload. The "lozenges" are drawn based on the payload state, and I can't reproduce the duplicate lozenges (or payload) on my end.

Thanks!

[Nic Hart]

I'll be happy to provide captures and logs but as the problem is easily and consistently reproducible it might be better for you (Stepstone) to replicate it. Just create a standard amazon aws windows instance (I've using their standard Win server 2008R2 + SQL express AMI for my tests). And go install Confluence+zen intranet. I've tested instances of capacity up to C2 extra large, with JVM memory settings to match, and have never seen the process complete in a single session. If it would help I'll contribute a step-by-step script for configuring the aws windows server ( not automated script though ;-( ).

[Darryl Duke]

Hi Nic,

Sounds like your Confluence might be running out of memory? Are you seeing any messages in the catalina.out or atlassian-confluence.log files? This page might help with the diagnosis and repair:

https://confluence.atlassian.com/display/CONFKB/How+to+Fix+Out+of+Memory+Errors+by+Increasing+Available+Memory

[Nic Hart]

Thanks Darryl but I doubt it.

I followed the advice you referred to and increased start/max mem to 1G/3G, permgen to 512M and threadsize max to 512k. (It's 4G VM). No real difference. The strange thing is - each time I restart the service I get 1 or 2 more green lozenges in the Intranet install task list. Finally I uninstalled all Zen inc Foundation, then resinstalled - and the part complete task list just came back (and did a couple more again).

I'm going to reinstall from zero.... it's a new site so no big deal, though not what you want in front of a customer