|
Description
|
customer has a system with Solaris 10 + 419 external disks.
For the 126th disk onwards, "mpathadm show lu <unit name>"
reported the following error.
----------
Error: Logical-unit 'unit name' is not found.
----------
For the disks before the 126th disk, "mpathadm show lu <unit name>" reponse was very slow.
They changed the limit to 256 descripters using the limit command.
And then they had an issue with the 256th lun.
The truss output (mpathadm.126.truss.log and 126.out ),open() reported "Err#24 EMFILE" error.
It seems that due to there are too many open files.
There are truss outputs ( 126.out and mpathadm.126.truss.log )
The first also shows it hitting a limit on open files.
The limit shows file descriptors as 256
cc092-A4[root]{5}# limit
cputime unlimited
filesize unlimited
datasize unlimited
stacksize 8192 kbytes
coredumpsize unlimited
descriptors 256
memorysize unlimited
And now they have set it as
acc092-A4[root]{6}# limit descriptors 1024
acc092-A4[root]{8}# limit
cputime unlimited
filesize unlimited¿
datasize unlimited¿
stacksize 8192 kbytes
coredumpsize unlimited
descriptors 1024
memorysize unlimited
Second truss shows:
21387/1: 4.3344 0.0001 open("/dev/rdsk/c10t6005076307FFC494000000000000
1573d0s2", O_RDONLY|O_NDELAY) = 255
and then
21387/1: 4.3396 0.0001 open("/dev/rdsk/c10t6005076307FFC494000000000000
1572d0s2", O_RDONLY|O_NDELAY) Err#24 EMFILE
There are also have these prior to the above open.
21387/1: 4.3389 0.0002 open("/var/run/syslog_door", O_RDONLY)
Err#24 EMFILE
Looks like open files require to be closed here.
|