[GE users] sge_qmaster heavy memory usage

Don Shesnicky dshesnicky at enqsemi.com
Mon Sep 27 19:22:30 BST 2004


 
I'm running 6.0 under redhat 7.2,  I have 10 sge_qmaster processes on my
master 
taking up 1.6 gig of memory. I found a bug report about running with
schedd_job_info
turned off. Is that the solution here as well? 

They appear to be all related according to ps -jax. Output from commands
listed
below.

Don

processes from top:

  PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME COMMAND
 1296 sgeadmin  15   0  163M 160M  2228 S     3.5  4.1  1024m
sge_qmaster
 1294 sgeadmin  13   0  163M 160M  2228 S     2.6  4.1 719:02
sge_qmaster
 1284 sgeadmin   9   0  163M 160M  2228 S     0.0  4.1   0:00
sge_qmaster
 1287 root       9   0  163M 160M  2228 S     0.0  4.1  15:14
sge_qmaster
 1288 root       9   0  163M 160M  2228 S     0.0  4.1   9:44
sge_qmaster
 1289 root       9   0  163M 160M  2228 S     0.0  4.1   0:14
sge_qmaster
 1290 root       9   0  163M 160M  2228 S     0.0  4.1 107:55
sge_qmaster
 1291 root       9   0  163M 160M  2228 S     0.0  4.1 160:37
sge_qmaster
 1293 sgeadmin   9   0  163M 160M  2228 S     0.0  4.1   6:20
sge_qmaster
 1295 sgeadmin   9   0  163M 160M  2228 S     0.0  4.1   0:00
sge_qmaster

pid/ppid list:

 PPID   PID  PGID   SID TTY      TPGID STAT   UID   TIME COMMAND
    1  1284  1283  1283 ?           -1 S     1087   0:00
/tools/sge/6.0/bin/lx24
 1284  1287  1283  1283 ?           -1 S        0  15:14
/tools/sge/6.0/bin/lx24
 1287  1288  1283  1283 ?           -1 S        0   9:44
/tools/sge/6.0/bin/lx24
 1287  1289  1283  1283 ?           -1 S        0   0:14
/tools/sge/6.0/bin/lx24
 1287  1290  1283  1283 ?           -1 S        0 107:55
/tools/sge/6.0/bin/lx24
 1287  1291  1283  1283 ?           -1 S        0 160:37
/tools/sge/6.0/bin/lx24
 1287  1293  1283  1283 ?           -1 S     1087   6:20
/tools/sge/6.0/bin/lx24
 1287  1294  1283  1283 ?           -1 S     1087 719:21
/tools/sge/6.0/bin/lx24
 1287  1295  1283  1283 ?           -1 S     1087   0:00
/tools/sge/6.0/bin/lx24
 1287  1296  1283  1283 ?           -1 S     1087 1025:17
/tools/sge/6.0/bin/lx2
    1  1299  1299   349 ?           -1 S     1087 748:30
/tools/sge/6.0/bin/lx24

bug report:

When submitting a constant stream of short running
jobs into Grid Engine with the schedd_job_info
switched on, scheduler and qmaster consume a huge
amount of memory (several 100 MB with only some
400 running or pending jobs in the system).

Workaround:
If schedd_job_info is switched off, the memory
requirement sinks to about 10% compared to the
above scenario.


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list