Cannot perform "Get Latest" on large repository

dwayne_davis · Post by **dwayne_davis** » Mon Jul 19, 2004 4:00 pm

I'm having trouble performing a "Get Latest" on a directory in my repository. This is also happening with a colleague of mine, on the same repository and folder.

When we do a "Get Latest", the process runs for several minutes with "Requesting files..." in the status bar. After several minutes, a message box pops up that says "The Vault server could not be contacted to perform this operation. Your network connection to the server may have been interrupted. Please verify your network settings using the Options dialig under the tools menu in the Vault GUI Client."

This doesn't seem to be a problem for smaller directories; the directory in question contains several hundred megabytes worth of data in various subdirectories.

I checked the server log file, which is currently in Debug mode; it contains nothing for these requests.

The Vault Session Timeout is set to the default of 4320 minutes; so no problem there. I checked the Vault.config and see the SqlCommandTimeout; it is set to 360, which I'm assuming is seconds. Should I up the timelimit on this?

Post by **jeremy_sg** » Tue Jul 20, 2004 7:28 am

Dwayne,

Since this is a network issue, it would be helpful to have more information about your connection. Are you going through a firewall to get to the server? It's likely that some network piece is timing out after a certain limit, and killing the connection. Does the get work if you use a client on te server machine connecting to localhost?

You are right to suspect SqlCommandTimeout, but it's doubtful that is the cause of the problem, as the server always logs Sql exceptions.

Could you post a snippet from the log file. The more you include, the better.

dan · Post by **dan** » Tue Jul 20, 2004 7:45 am

Another thing to test: Is the server restarting after your Get? If so, and you are getting hundreds of MBs of data, it might be that IIS is using up too much RAM and recycling itself. Check the event log to verify.

dwayne_davis · Post by **dwayne_davis** » Tue Jul 20, 2004 3:54 pm

I have spoken with our IT people to ensure that we are not going through a firewall to get to the server. We did recently make some changes to our network (on the 9th of July), but we've never had a firewall between our clients and the vault server.

I tried a 'get latest' on the server. It took a long time - long enough that I had to lock the screen and come back to it. When I did, there was nothing on the screen to indicate a failure, but I didn't get any files either - just the directory structure (tried on an empty directory).

This has become a critical issue for our group - our Source Control is unusable. We cannot reliable retrieve data from the repository.

We did a test on another repository to ensure that this is not something specific with the repository in question. We tried to upload a large amount of files (various binary and nonbinary files totally over 200 MB worth of data). This failed as well stating that the server was unavailable for transaction begin.

At one point, trying to do a get latest of our repository, the process reached 66% of retrieving files, hung for some time the output to the message log "An error occurred reading from the download stream ("The operation has timed-out."), retrying...". It then retrieved files, but I noticed that it was looping through the same set of files over and over again (I could tell by watching the status bar output). I aborted, and then it output the list of files retrieved and "Finished get latest of $/Network Based Gaming". I did not check the intergrity of the files retrieved; deleted the directory and tried again, I got the same error from my original post. See the "FailedGet.txt" attachment for details of this particular failure. Note especially that some of the files are duplicated in this output about 13 times. I can email this file to you, if you like. I'd rather not post it as it is large and lists all of the files we've been working on.

I've attached the sgvault.log file (as txt to not upset the forum). There doesn't seem to be any usefull data in here, and this is at the Debug level. I have already tried shutting down and restarting the IIS process. There is nothing in the event log to indicate that IIS is having problems.

I should clarify that I'm on the 2.0.3 version of client and server. I have another bug I posted on the 9th where rollback was failing, http://support.sourcegear.com/viewtopic.php?t=1403. Note that although this problem exhibited itself on the 9th (the same day as our network changes), the problem occurred several hours before the network change. I don't know if the rollback issue is related to the current problem or not. I only know that Vault is currently unusable to us.

We're looking to put the repository on another database, to rule out issues hitting the database. But, there is nothing in the log about the database being an issue, so we're really grasping at straws. However, since we can't use it anyways it seems worth a try.

Post by **jeremy_sg** » Wed Jul 21, 2004 6:52 am

Dwayne,

Send me an email with your phone number, and we'll work offline to fix this.

-Jeremy

Post by **jeremy_sg** » Mon Aug 09, 2004 10:49 am

This was revealed to be a network problem. Moving the Vault server to a different machine corrected the problem. This also solved dwayne's other issue:

http://support.sourcegear.com/viewtopic.php?t=1403