Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-30 Thread Grant Ingersoll
For the record, the test was meant to test the new partial optimize capabilities by counting the number of segments available after doing partial optimizes. -Grant On Jul 29, 2008, at 5:58 PM, Tricia Williams wrote: Works for me! I don't really understand the purpose of the test, but it

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Tricia Williams
This same thing happens to me since DirectUpdateHandlerOptimizeTest was added to the repository. How does one increase the file handle limit in ubuntu? Thanks, Tricia Shalin Shekhar Mangar wrote: Yes, it happens on a fresh checkout too. cat /proc/sys/fs/file-max gives 204979 on my box. The

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Shalin Shekhar Mangar
You can increase it through /etc/security/limits.conf in ubuntu On Wed, Jul 30, 2008 at 1:09 AM, Tricia Williams [EMAIL PROTECTED]wrote: This same thing happens to me since DirectUpdateHandlerOptimizeTest was added to the repository. How does one increase the file handle limit in ubuntu?

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Yonik Seeley
I just committed a fix that will make the test use the compound file format. Hopefully that will be sufficient. -Yonik On Tue, Jul 29, 2008 at 3:39 PM, Tricia Williams [EMAIL PROTECTED] wrote: This same thing happens to me since DirectUpdateHandlerOptimizeTest was added to the repository.

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Tricia Williams
As of revision 680834 the DirectUpdateHandlerOptimizeTest is still failing. I haven't made any changes to the file handle limit on my machine. Tricia Yonik Seeley wrote: I just committed a fix that will make the test use the compound file format. Hopefully that will be sufficient. -Yonik

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Yonik Seeley
OK, I've now scaled back the test by a factor of 10 (50 segments instead of 500). -Yonik On Tue, Jul 29, 2008 at 4:53 PM, Tricia Williams [EMAIL PROTECTED] wrote: As of revision 680834 the DirectUpdateHandlerOptimizeTest is still failing. I haven't made any changes to the file handle limit on

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Tricia Williams
Works for me! I don't really understand the purpose of the test, but it looks like it is meant as a weak stress test. If the test is significantly scaled back, is the test still accomplishing what it is mean to do? Should the onus instead be on the developer to meet a minimum requirement?

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-29 Thread Shalin Shekhar Mangar
Works for me too. Thanks Yonik! On Wed, Jul 30, 2008 at 3:28 AM, Tricia Williams [EMAIL PROTECTED]wrote: Works for me! I don't really understand the purpose of the test, but it looks like it is meant as a weak stress test. If the test is significantly scaled back, is the test still

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-21 Thread Shalin Shekhar Mangar
Yes, it happens on a fresh checkout too. cat /proc/sys/fs/file-max gives 204979 on my box. The test which fails is testOptimize. I increased the file handle limit to 4096 but it still fails. Then I increased it to 16384 and the test passed. On Mon, Jul 21, 2008 at 2:58 AM, Yonik Seeley [EMAIL

Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-20 Thread Shalin Shekhar Mangar
Hi, The DirectUpdateHandlerOptimizeTest fails on my local box due to too many open files. The nightly build does not fail so I'm assuming it must be something specific to my setup. Has anyone else seen this problem on their local environment? ulimit -n gives 1024 on my Ubuntu Hardy Heron.

Re: Too many open files with DirectUpdateHandlerOptimizeTest

2008-07-20 Thread Yonik Seeley
Does it happen on a fresh checkout? I use windows (and cygwin for the familiar command line interface), which has higher limits. Also watch out for the system-wide limit: http://www.cs.wisc.edu/condor/condorg/linux_scalability.html If it happens with a fresh checkout, perhaps we could lower the