Community technical support mailing list was retired 2010 and replaced with a professional technical support team. For assistance please contact: Pre-sales Technical support via email to sales@march-hare.com.
> From: Harrison, Andy > It does seem to be load related - the more people trying to access the > server, the more likely it is to happen. However, the CPU load doesn't > seem to be very much, it's just the disk that's hard at work. > Last time > I checked, there were 10 active cvs.exe processes, but only > 2% CPU load. Ouch. That means something else is the bottleneck (or there's a gotcha in the code, but let's be optimistic and assume load-related for the moment), and it sounds like disk. Which could also be swap-related if there's excessive memory use - ISTR Tony saying cvs.exe could grow quite large. Any chance you could monitor disk subsystem load (the simplest 'single number' is average disk queue length on the phyical disk(s) that have the CVS repo on, and also on the CVS temp dir if that's different)? Also Memory/pages per second? When it happens again, those values could be interesting. Any chance you can add a separate spindle or set of spindles for CVS temp if you haven't already? Might make some difference, might make none at all. I assume there are no I/O timeouts or similar listed in the event log? What have I missed? :-) - Peter