memory,disk,cpu process.....

[Rushank]

Hello,

How to assess the data from Glance or sar or swapinfo or vmstat or.
For example
N-class HP 11 Physical Mem 8G with Oracle 8.x

dbc_max_pct=5
dbc_min_pct=5

Is this parameter OK..?

2.
swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RES PRI NAME
dev 1500 10 1490 1% 0 - 1 /dev/vg00/lvol2
reserve - 803 -803
memory 788 240 548 30%
total 2288 1053 1235 46% - 0 -

Is this swap memory OK Or If I need more swap What would be steps, pseudoswap is configured

How can I check Is there any memory leak from Glance or sar.

In a nutshell, How do I check a Server whether it's performing well or It needs any tuning

Thanks in Advance

[harry d brown jr]

Id set the min to 5% and the max to 10%.

As for performance, it's a "reality" game: What kind of performance do you expect from what you have?

Each application has it's own requirements, and as multiple applications are placed on a single server the more complicated things get. The best thing to do, in my opinion is to purchase perfview and do a TREND ANALYSIS. Without a TA you have no pivot point to base any assumtions on.

live free or die
harry

[Rushank]

How do I analyze the perfview or glance data how to perform Trend Analysis.

What about swap and memory leak??

[Craig Rants]

Rusank,
HP has a product add on for Perfview called something like Trend Analyzer.

As far as memory leaks, I don't know of anything that detects those.

Good Luck,
C

[Roger Baptiste]

<>

It depends on what you are looking for. For instance on glance, the immediate things to look is the CPU, Memory and IO usage. That is the broadest categorisation of bottlenecks. In swapinfo, look at how much swap is being consumed!
Sar has so many options, in
the simplest one
sar 5 5 shows the cpu usage
sar -b shows the buffer cache usage. sar -d shows the disk channel usage (higher service times means bottlenecks on those disks/channels)


<dbc_min_pct=5 >>

This falls under kernel configuration tuning. You can look up some of your kernel parms usage through
glance ->t which shows the % of usage.
As far as dbc parm is concerned make the max as 10% as there needs to be some difference between max and min to give it a range.


<>

46% overall and 1% of diskswap is fine.


<>

Configure measureware/perfview and constantly monitor the historical usage of the parameters. perfview is a very good tool to use and a nice place to start off with.

HTH
raj

[Rushank]

Thanks Raj,

Is the swapsize is OK..?? Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1024 0 1024 0% 0 - 1 /dev/vg00/lvol2
reserve - 675 -675
memory 6278 1485 4793 24%
total 7302 2160 5142 30% - 0 -

[Roger Baptiste]

<TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1024 0 1024 0% 0 - 1 /dev/vg00/lvol2
reserve - 675 -675
memory 6278 1485 4793 24%
total 7302 2160 5142 30% - 0 ->>

I understand , this is on a 8GB RAM system. Your total swap including pseudo-swap is around 7 Gb. Normally, Swap should be atleast equal to the size of the RAM, which is 8Gb here. But, the swap usage seems low in the output you posted. If this is the case even during the system busy time, then you dont need to change it. If the swap usage increase during busy period, then add another device swap of 1Gb.

HTH
raj

[Eileen Millen]

We have EMC disk on our Oracle system.
The EMC has 1.5Gb of cache.
Oracle recommended that we set dbc_max_pct to
5, because of the EMC cache and Oracle handling it's own caching.

Eileen