[cvsnt] Re: Performance Woes...

[Jeremy Weber]

Just an FYI...

I was able to fix my issues!  Whohooo....

I did so many things that I cant be sure what exactly fixed the issue.... 
Once I realized that file copy operations to the server in question were way 
slow compared to our production box, I had a good indication that something 
was wrong with the box and not necessarily CVS.  The final two fixes I did 
were with the windows registry.   I disabled SMB Signing and added a missing 
registry key for Opportunistic Locking.

Thanks for everyone's responses...



"Matt Schuckmann" <matthew_schuckmann at amat.com> wrote in message 
news:d87m5m$nsp$1 at paris.nodomain.org...
> Jeremy I'm in sort of the same boat as you, we have to run on win server
> 2003 and we've got a few binary files that are really slowing things down.
> Below is an observation that I made and I posted it to this list a while
> back but got no response.
>
> I have noticed that CVSNT can be very slow with large repository binary
> files.
>
> I have a 400mb repository file (the ,v file was 400 mb, the checked out 
> file
> was about 12mb). CVSNT takes nearly an hour to perform a simple operation 
> on
> this file, not checkout something simple like check for the existence of a
> tag.
>
> What I noticed is that the entire file appeared to be getting loaded into
> memory on the server (the memory usage on the server went up by 400mb, but
> very slowly). It appears that there wasn't enough memory so it started
> thrashing on the virtual memory and thus the disk. It was just the loading
> of the file into memory took such a long time, once it was done sucking up
> the memory the actuall operation was very quick.
>
> This was on my test server, a P3 700mhz with only a couple hundred megs of
> ram, Win2k and a very recient version of CVSNT.
>
> So far I've just ignored the problem until I absolutely have to deal with 
> it
> at which time I don't what I'll do, probably delete some unneeded 
> revisions
> (does that work for binary files?) I'm hoping that my production server 
> with
> more memory will handle it better.
>
> I have no idea how a vanila version of CVS would perform on this file.
>
>
>
> "Jeremy Weber" <jeremy.weber at active-endpoints.com> wrote in message
> news:d87919$hse$1 at paris.nodomain.org...
>> Ughhh, I am having no success in trying to upgrade CVS in terms of
>> performance.  I would like to move our existing respository to Windows
> 2003
>> Server Standard and its a headache.
>>
>> Our current production system is as follows:
>>
>> - CVSNT - CVS Version: Concurrent Versions System (CVS) 1.10.8
>> (client/server)
>> - Windows 2003 Enterprise Server - Runs DHCP, TS, DNS, PDC, NAV 9.0 - not
>> running on accessed files.
>> - Xeon 2.8, 1 GB Ram, 2 UATA 7200rpm 40 GB Drives....
>>
>> When I checkout our lib directory which contains binary files via pserver
>> over out lan, it takes 56 seconds.
>>
>> New Test CVS Server. ( Dual Booting Windows 2003 Standard/RedHat ES 4.0)
>>
>> On Drive 1:
>> - Windows 2003 Standard Server - Nothing installed/running but CVS.
>> - CVSNT -  CVS Version: Concurrent Versions System (CVS) 1.10.8
>> (client/server).
>> -  2 Xeon 2.8, 2 GB RAM, 1 SATA 7200rpm 80 GB Drive
>> On Drive 2:
>> - RedHat Enterprise 4.0
>> - CVS - Concurrent Versions System (CVS) 1.11.17 (client/server)
>> - 2 Xeon 2.8, 2 GB RAM, 1 SATA 7200rpm 80 GB Drive
>>
>>
>> Drive 1 Tests
>> --------------
>> Test 1 - Checkout time: 24 minutes
>> Test 2 - Disabled HTing, and CPU2 - 3.5 minutes
>>
>> Some other tests I have run:
>>     - local v. domain users
>>     - pserver v sspi
>>     - creating cvs temp on a ram disk
>>     - setting cvsroot with an ip/hostname
>>     - Using CVSNT  (2.5.01.1976)
>>
>>
>> Drive 2 Tests
>> --------------
>> Test1 - Checkout Time: 15 seconds
>>
>> Obviously I have some issues with performance.... 15 seconds would be
> ideal,
>> 24 minutes is utterly unacceptable.   Im no network administrator but 
>> have
>> run the Event Viewer and Performance monitoring tools on 2003 and nothing
>> glaring stands out.    I have read all posts on this group that I could
> find
>> regarding performance and suggestions have worked, at least enough for me
> to
>> cut over to the new server.
>>
>> I'd be willing to test code, implement ideas, anything at this point. Can
>> anyone help me?
>>
>>
>> Thanks
>>
>> J
>>
>>
>>
>>
>>
>
>