We are having unbelievable problems, apparently with App Pools on three of our 5 15.0.3 and/or 15.1 clients and are looking for hints/advise/suggestions. One installation in particular. This particular customer is getting very frustrated at the amount of time it takes, one, to log in via the desktop, and two moving between tabs. When we look at the event log, we see literally hundreds of ASI errors and warnings in a 24 hour period of time. TS has suggested changing this setting and that setting and increasing the worker proceses, etc, and NOTHING seems to help. Increasing the worker processes just consumes huge amounts of system resources. CPU runs at 100% almost constantly and memory runs at around 91% constantly. the drives are thrashing because of the contstant memory swapping. Adding memory doesn't seem to do much and the machine meets or exceeds the minimum requirements in every way. we had this customer, originally, hosted on the same server as one other iMIS client running two instances of iMIS, and have since moved them off on to their own server. This is a very small client - 4 user, and we haven't even dared take their public site live yet. Tthey are on 15.1.1 latest patch. Other clients have similar issues but not to this extent. This is very frustrating for all involved. I cannot beging to count the number of non-billable hours we have spent on this issue, and the client is starting to laugh at the "customer for life" moto ASI has posted on the startup page. Can anyone offer any advise before we loose this customer and our minds?
App Pool issues causing major headaches
this is server 2008 - does
this is server 2008 - does that matter? I know there are some differences between iis6 and iis7 that are part of my "learning curve".
iMIS Docs has this
Yeah, i have been through
Yeah, i have been through all of those. I am about ready to give up and go get a job at McDonalds....
this may have no relation at
this may have no relation at all, but as I read 100% CPU I think 'Autolog out after' in File > System setup. Make sure this option is not activated.
thanks, I will double check
thanks, I will double check that. We did see that Microsoft Error Reporting (cw20.exe) runs every three or four minutes and consumes every ounce of available cpu when it fires off so we have crippled that and that seems to have helpd with the CPU hits.
Multiple worker processes = bad
My experience has been that iMIS isn't stable when you turn multiple worker processes above 1. I think I even heard an ASI presenter say not to use multiple processes because iMIS wasn't designed for it.
We've also found some improvements by making sure the different sites use different pooled SQL connections. Any difference in the connection string will cause .NET to pool connections differently. We got relief for one client by adding "Max Pool Size=200;Connection Lifetime=3600;" to one site, and "Max Pool Size=201;Connection Lifetime=3600;" to the other. This guaranteed that Public user connections wouldn't prevent Casual site access. Max Pool Size tells .NET how many simultaneous SQL connections it can have, and Connection Lifetime means .NET will recycle any SQL connection over N seconds (6 hours in our case).
-- Bruce
thanks Bruce. Nice to see we
thanks Bruce. Nice to see we are not loosing our minds here and there are others experiencing some of the same issues. This stuff is actually starting to make some sense to me.
Would you mind telling me
Would you mind telling me what the configuration of the server is that you refer to above? Specifically number of users, amount of memory, cpu? Do you run the app on the same server as the database? Just curious.
Server specs
Here's the specs.
- Dual 3Ghz Xeon
- 2GB RAM
- Server hosts both casual and public sites
- Server is self-hosted, located in a DMZ at the client's offices.
- SQL is on a separate server inside the LAN (behind the second firewall). Very capable.
- Staff view is a different domain name and different site than the public site. (E.G. http://staff.example.org/imis with no public IP and http://logon.example.org/public with a public IP and published DNS)
- The web server hosts about 6 other mostly-static websites.
- About 20 staff users, unlimited public users
- The performance problem was in iPV events, then version 15.0.1, and a globally-known guest speaker.
-- Bruce
Hmmmmm...... TS is telling
Hmmmmm...... TS is telling us that we need to add memory to our existing 4Gb on this server (Dual 2Ghz Xenon) for a 4 user system that is still not released for public access yet....... The only difference is that the time being, the DB is on the same server (Sql Express 2005).
Not to contradict Tech Support...
4GB should be plenty, unless you have a lot of other websites and services running on the same box. You can achieve the same thing for now by setting a reasonable upper limit for SQL Express. Max RAM of about 20% of all your active databases should be a comfortable number that won't squeeze SQL too much.
Some tuning steps will improve one area and hurt another. For instance, setting your app pool to recycle more frequently may cause an increase in the number of "startup" bug problems.
Check your disk space -- every drive should have at least 15% free or performance will tank.
--
Bruce Wilson
Director, Technical Services
RSM McGladrey, Inc.
Multiple worker processes = bad
Can you tell me where you are modifying these settings? I am having a lot of the same problems with a 15.1 site. Thanks
Thanks, Bruce. I hadn't
Thanks, Bruce. I hadn't thought of setting the max limit on SQL Express. This issue has maybe been "resolved" - at least I hope. The issue has finally been reproducible and flagged as a high level defect.
Try these steps
http://support.microsoft.com/kb/816517