G'day,
I think Continua build 1633 has introduced a bug. If I look at the build times on a simple build and run unit tests project, they went from 1 minute to having me to manually stop them close to an hour later. If you look at the duration log for work I been doing from lunch time today on a project, you can see that the time has blown out and I have to manually stop the build.
Attachment unavailable
What build were you using previously? Also, what does the log for those builds show?
The first set of values were triggered from v1.0.0.1610, while the builds I had to manually stop were v1.0.0.1633.
The log looks normal, but I suspect the hook which receives the "finished" event for the UI isn't working correctly. As you can see by the log, it finished fairly quickly.
Attachment unavailable
Hi Jamie
Very strange. Can you show the Timeline page (just the last few entries).
Vincent,
Here is build 33, the last fast build
Attachment unavailable
I've been getting this a lot too, I think before 1633 too.
* Log says everything complete fine.
* Time-line says agent is done with the first stage. Doesn't start the second.
* Stage percentage bar is full, but still blue (not green to say it's completed).
Can you check if there are any errors reported in the Continua & Windows Event logs? We’ve not been able to reproduce this here so far.
Also can you confirm your agents are up to date?
Hi Guys
When you see this issue, were the builds started by a trigger or manually?
Just migrated to 1633 for server and developer workstation agent
Both agents running the same version.
Repository change triggered the build
Build ran on developer workstation agent. (win 8 x64)
Server event log clear. (win server 2012 x 64)
Developer workstation agent event log had lots of during the time of failure.
“taskhostex (5092) The database cache size maintenance task has taken 75 seconds without completing. This may result in severe performance degradation. Current cache size is 21 buffers above the configured cache limit (111 percent of target). Cache size maintenance evicted 6 buffers, made 8806 flush attempts, and successfully flushed 0 buffers. It has run 7338 times since maintenance was triggered.”
I’ve done the usual M$ tricks for the OS and rebooted, bla, bla, bla. Will see if I can force it to happen again with out any other moving software components (aka service packs etc). I just wonder if the Windows 8 2 day reboot might be part of the issue. I know I did reboot/upgrade earlier in the day, just not sure about the Important service packs from MS which I triggered manually late on Friday.
I don’t think that error is related to continua, from what I can tell, taskhostex is a host application for dll’s.
The strange thing is the stage is actually completing on the agent, the last entry in the timeline is the last message sent from the agent to the server after a stage has complete. The very next line of code is a callback from the agent to the server to tell it the stage has completed. One of the guys here did see this issue on Friday, but we have not been able to reproduce it since…
I’m happy to call it a Friday “Feature”.
Nope, it’s a bug… and it’s been there for a while, it’s a timing bug.
We were able to reproduce it on multiple dev machines (not reliably), and with some extra debug information we proved my theory (messages still in the queue when the agent callback proxy was being disposed). New build after lunch (just investigating another issue which I hope will be easier to track down). Thanks for your help.
New build with the fix is uploaded (1.0.0.1647)