Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations IamaSherpa on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Fast connect and paging space problem

Status
Not open for further replies.

henrik

IS-IT--Management
Jul 4, 2000
6
DK
I have a problem on a AIX running Fast connect where the server has begun to have slow performance when more users are connected. At the beginning we had a H70 running mirrored SSA, 512MB memory and a swap space on 768MB.

When more users were connected (below 50) the errorlog from time to time reported paging problems (cifsserver & cifsprintserver) so we created a second swap space of 512MB but the performance is not regained!

My question is if any body have had the same experience and how you handled it?

If am thinking of using the vmtune command to see if that could help but.....?
 
Henrik,

What paging problems are you seeing?

Could you try running 'vmstat 1' both while the system's behaving itself and while it isn't?

One of the interesting figures is the 'po' (Page Out) column, would you post about ten lines from when the system is ok and the same from when it is not?
Mike
michael.j.lacey@ntlworld.com
 
i suggest you increase your paging space to double your physical memory (1024)and yes ou can do performance testing as MIkeLacey has said vmstat or iostat this will give you an ida of what is happening on the machine .

hope it works
 
Real memory is eaten - a second paging space has allready been added - resulting in some extra hours before Fast Connect had to be restarted! Real memory has been reduced as well (vmtune) - downto 40% (from 80%) - resulting in problems much before. There is only 10 users, 10 shares and one printshare in Fast Connect!! vmstat gives you (at the beginning) plenty idle% cpu but at the end about 90% is allocated to sys and at the same time a lot of (500) CifsPrintServer can be notified!
 
Henrik,

Is the process size of Fast Connect increasing? If it is that could well point to a problem with Fast Connect itself. (perhaps a memory leak, memory being alocated but not being released when its finished with - not something *you* can correct - you would need to report that to whomever you bought Fast Connect from.

You can find out using the ps command, the field you need to look at is RSS (Resident Set Size)

Mike
michael.j.lacey@ntlworld.com
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top