GLANCE QUESTION

[Nobody's Hero]

I'm monitoring an L2000, 2 440's and 2 gig of memory. System is running VP/O and CiscoWorks 2000. Both seem to require a good bit of resources. In glance plus, memory graph. I noticed a constant amount of 'deactivation rate' while the 'wait queue` is showing a low flat line. Every once in a while the wait queue will spike fo a short period. What is this telling me. Do I need more swap area? Can someone explain the 'deactivation rate'?

Thanks,
Bob Menefee

[George_Dodds · Accepted solution]

There's a brief explanation of the last reply on the following link

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x1eb1d5fab40ed6118ff40090279cd0f9,00.html

Cheers

George

[Krishna Prasad]

It sounds like you need more swap.

What does swapinfo -ta tell you.

And when in glance type m to look at memory how much free memory do you have?

[Craig Rants]

Deactivation is a bad thing. I was told by an HP ASE that basically a process is told not to run, thus deactivated. I probably did not do his explanation proper justice, but deactivations are a key indicator of some type of memory problem. Now being a consistent rate is not as important as the total number of deactivations represented by the current and cummulative rates. I would be concerned if they started getting over 1.0, but that is just knowing my servers, this may normal for yours. Others may provide a different view.

Hope this helps,
Craig

[harry d brown jr]

Robert,

I'm impressed that you can get VPO and Ciscoworks running in 2GB of memory. I threw an N-class server with 4 cpu's and 6GB of memory at it. I'm wondering if I overkilled the situation?

What does your memory report show (not the graph)? How much buffer cache do you have allocated? I'm thinking you need more memory.

live free or die
harry

[Nobody's Hero]

Harry,
Using memory report. Shows 435.1mb. What are you thinking? AM I in bad shape?

[Nobody's Hero]

Ron swapinfo -tm, total = 55%

[Craig Rants]

Robert, what type of database do you have supporting VPO? Oracle, or are you using flat files? The tough part of using both apps on one box is that Ciscoworks includes a Sybase database, thus if you also have Oracle on there the SGA size is probably quite large. Run ipcs -a to find out if you have a large SGA size, that could provide some valueable information for you.

GL,
C

[Nobody's Hero]

Thanks Craig, I'm running Oracle 816. Seems to playing well with sybase. I ran ipcs -a, not sure what I am lookin at. Sort of a novice with memory management. How do I see my SGA value?

[Bill Hassell]

Deactivation is done when there is memory pressure (need more RAM, not enough available). A process may be deactivated that is alreaday sleeping or waiting on I/O. As a metric, deactivations isn't that useful. Adding swap space does nothing unless you have already run out.

What matters is the page-out rate. These are real movements of data from RAM to disk. Paging essentially removes pieces of deactivated programs (more efficient than full program swaps) and this rate is measured by page-out. vmstat, Glance and gpm all show page-outs. Page-in isn't useful because it shows program startups as well as return from swap.

[Krishna Prasad]

What might be happening then is that one of your database's allocated memory may be full. I.E. your Oracle SGA may be to small and when more information is needed it allocates the memory out. If you have 400+mb of memory free you can give more memory to databases.