[SunRescue] Loads

Peter Koch koch at pz.pirmasens.de
Wed Mar 8 02:49:24 CST 2000


Hi!

>4:56pm  up 28 day(s),  8:38,  450 users,  load average: 1.32, 1.60, 1.64
>(SS10, 4 HyperSPARC 95MHz)

With 450 current users, this is a damned light load.

>4:59pm  up 28 day(s),  8:42,  448 users,  load average: 2.18, 1.65, 1.62

Aha. Do the users complain? No!?!
The machine has 4 processors, so a load below 4 is acceptable.
If you have periods over minutes/hours with a load above 4
it becomes critical.

>has 512MB of RAM and has about <10MB left...

This is normal. The machine will use all free memory as file
system cache and reclaim memory as needed. The 10 MB are
a reserve pool for short period demands.
How much swap space usage do you have (try "swap -s")?
How is your swap space arranged? Do you have several
swap partitions on several independent disks (good)
or do you have a single swap file (bad)?

Have you tried to run "top" during the lag periods? Maybe
you can identify abusal processes/programs/users...

One thing: 450 users is a lot. Maybe too much. What are
these users doing? Reading news and mail? That would be
fine. Doing compiler or database jobs? That would be
terrible (even if only a handful of such users exists).
Remove such users or the database/compilers from the
machine.

Since you cannot upgrade the machine anymore (512 MB
RAM is the maximum and four 95MHz Hypersparc are surely
the absolute maximum the mainboard can support) you need
to reduce the load. Or buy a new machine :-(

Tschuess

Peter

P.S.: Sun doesn't recommend more than 2 processors in
the SS20 that are faster than 50 MHz. Maybe the
four Hypersparcs are clogging the memory bus...






More information about the rescue mailing list