[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Warnings in 7.4.2 not seen in version 7.2.x

On 07/20/2010 02:51 AM, James Osborne wrote:
> Dear All
> With a recent upgrade of a Condor submit server from 7.2.x to 7.4.2 I have 
> started to see some interesting warnings.
> That submit server mounts an NFS volume for its home partition and both I 
> and my users are now seeing warnings like
> WARNING: Can't determine whether log file /home/username/SWUS/job-log.txt 
> is on NFS
> Is there a configuration variable I can tweak to get Condor to ignore this 
> particular warning ?
> Thanks in advance
> James

It looks like there's no param to quiet that warning.


You should only be getting that warning if the statfs or statvfs calls are failing.

Will you first check and see if debug output says anything about "detect_nfs", e.g. "statfs(/home/username/SWUS/job-log.txt) failed: ???/???"?


Try: echo "cmd=/bin/true\nlog=/home/username/SWUS/job-log.txt\nqueue" | _CONDOR_TOOL_DEBUG=D_ALWAYS condor_submit -debug -dump /dev/null

If the debug output doesn't do the trick, will you include information about what binaries you are using, on which platforms, and maybe information of what statfs returns when pointed at /home/username/SWUS? See the statfs.c below. Seems like it would be useful to ship a diagnostic tool to gather such information. I wonder if the VDT has one.



$ cat statfs.c
#include <stdio.h>
#include <sys/vfs.h>

main(int argc, char **argv)
	struct statfs buf;
	if (statfs(argv[1], &buf)) {
		return 1;

	printf("f_type(%s) = 0x%X\n", argv[1], buf.f_type);
	return 0;
$ gcc statfs.c 
$ ./a.out /tmp
f_type(/tmp) = 0xEF53
$ man 2 statfs | grep 0xEF53
              EXT2_SUPER_MAGIC      0xEF53
              EXT3_SUPER_MAGIC      0xEF53