OpenEdge 10.1C02 performance

[oldemanw]

Dear All,

Does anyone have experience with Progress OE10.1C02 vs. OE10.1B05 on Linux?
We had to upgrade, because B05 could leave transactions open under certain circumstances.
Another thing that was changed was our operating system. We were on RHEL 3 and had to upgrade to RHEL 5.3 (because of 10.1C02).
Since then, we notice a significant performance drop. Any clue?

Previously, the machine would have a load between 0.6 and 1. Now we have an average load around 3~4, sometimes even peaking up to 12.

Hints are welcome. If I need to provide more information, please let me know.


Thanks in advance,
Willem

 

[TomBascom]

No, I haven't seen anything quite like that. Usually when something like this happens it is due to a parameter or configuration change. Sometimes due to a parameter default value change that Progress makes (I don't know of one between 10.1B and 10.1C though -- are you sure that you've got your versions right? I don't think that there is a 10.1B05...) The most recent significant default value change that I can think of was to -tmpbsize.

You should double check all of your startup parameters and configuration options -- maybe something was accidentally changed somewhere along the line.

IMHO load average isn't a very reliable indicator of much of anything. Far more interesting would be Progress centric metrics. The PROMON summary screen is a good place to start or you could download ProTop. You should complement that with more specific OS level tools to determine where the excess load is coming from. "top" is usually included with Linux butpersonally I find "nmon" to be a lot more useful. It has some really good disk utilization screens.

All of this would be much easier if you had been collecting historical performance and configuration data so that we could compare before and after. Without that we're mostly guessing.

 

[oldemanw]

Hi Tom,

Thanks for suggesting nmon, I'll ask our sysadmin to get it for me.

You're right about the B05 version, it was actually B03 :blush:. And yes, it's a pity I didn't collect historical data. Until B03 there seemed no reason to monitor performance, I just had a look every now and then. I better keep track now

Another thing caught my eye (using protop -thanks for that-), some extents are > 100% full, meaning the variable extents are in use.
Funny, I resized them a few months ago to cope with this year.
I'll sniff around a bit more using protop and double check our startup parameters.

Thanks,
Willem