Hello all,
I added a clause to my script.
sysinfo:::
/self->traceme==1 && pid == $1/
{
trace(execname);
printf("sysinfo: timestamp : %d" , timestamp);
}
A subsequent trace created a file of about 19000 lines.
I loaded in Excel to be able to subtrace timestamps etc.
The longest jump in timestamp is between the first pair of savectx and
restorectx at line 70.
I count 50 savectx/restorectx calls in my trace.
But the actual physical write as indicated by the sysinfo is almost at the end
of the file directly after the ufs_write and fop_write call end (line 18746).
24 <- tsd_agent_get oracle timestamp
1795469946839100
24 -> ufs_lockfs_top_vop_return oracle timestamp
1795469946841500
24 <- ufs_lockfs_top_vop_return oracle timestamp
1795469946844300
24 <- ufs_lockfs_end oracle timestamp
1795469946846700
24 <- ufs_write oracle timestamp
1795469946849600
24 <- fop_write oracle timestamp
1795469946853500 57,365,700
24 | pwrite syswrite oracle sysinfo timestamp
1795469946856800
24 | pwrite writech oracle sysinfo timestamp
1795469946860200
So it seems that the actual write not the bottle neck but
I attached a zip file with the excel document.
Am I right in thinking that is is more an OS issue than a storage issue?
Regards Hans-Peter
--
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
