Thanks to the following who responded to my original question, regarding the
slowness of the sync command on a SS1000E, which was locking up a Progress
database. (Full message included below)
"Jurgen M." <sysjxm@devetir.qld.gov.au>
Peter Marelas <maral@phase-one.com.au>
vqh@dwrock.dw.lucent.com (D461-Viet_Q_Hoang(0)82572)
Brion Leary <bleary@state.ma.us>
The problem has been partially resolved by two measures 1) adding NVRAM to
the system and 2) Doing a complete reinstall/upgrade to Solaris 2.5.1. Both
actions individually seemed to have a positive effect.
To further resolve the problem we will be looking at hardware upgrades -
perhaps Ultra's with the newer fast-wide disk packs.
Thanks again to those who provided useful insights.
-- Stephen Tremain Unix Systems Administrator Deutsche Bank Australia Email: stephent@bain.oz.au Phone: 61-2-258-1907Original Message:
> Dear All, > > We have been experiencing problems with a Progress data base application > running on an SS1000E. The user frontend hangs every 30 seconds for up to > 10 - 15 seconds. The problem has been traced to the user process calling > sync at every database checkpoint, and freezing to the sync returns. The > Progress version is 7.3C. (I have been told that Version 8 works differently > but that is not an option yet.) The Progress DBA's have been able to change > the frequency of the checkpoint to every 2 minutes, but then the system freezes > for 20-30 seconds. > > The time to sync the disks seems particularly slow on this machine, and > I am curious if any other admins also have this problem, or know of any > solutions. > > We have been tuning kernel parameters without much sucess. Curently > /etc/system contains: > > set tune_t_fsflushr = 50 > set autoup = 300 > set bufhwm = 10240 > > The following is some data gathered from our problem system (yeats) > and the most similar other system we have (spike - a sybase server) > > Yeats Spike > ----- ----- > System 1000E 1000E > No. CPU 4 x 85Mhz 6 x 85Mhz > OS 2.5 2.5.1 > Mem 1024M 416M > Disk 71.4 Gb 90.3 Gb > > Notes: > System load is similar, but yeats is the busier. > Spike has NVRAM, yeats does not. > Spike is mostly raw partitions, yeats is all FS > > > However, the time taken to "sync" the disks is vastly different. Over 10x > slower. > > spike: ~> /bin/time sync; /bin/time sync > > real 0.8 > user 0.0 > sys 0.4 > > real 0.6 > user 0.0 > sys 0.4 > spike: ~> > > > yeats: /> /bin/time sync; /bin/time sync > > real 10.4 > user 0.0 > sys 6.3 > > real 6.3 > user 0.0 > sys 5.6 > yeats: /> >
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:15 CDT