Backpressure? High Memory Utilization

[hunterdw]

Exchange 2007 SP1 Update Rollup 4 applied.
Virtual Machine in an ESX VI Cluster
2 vCPUs
16gig ram
24gig-48gig managed pagefile
400 users
about 500gig of data in 12 mailstores

I'm starting to see Event Id 1009 / 15004 / 15005 in my event logs

I've read

http://technet.microsoft.com/en-us/library/bb201658(EXCHG.80).aspx

and while it gives a great "background" it doesn't have great "troubleshooting" features.

When 15004 gets to "medium" and "high" then my mailflow stops... and then my cell phone blows up


Can anyone offer some pointers?

Thanks,

--DW

 

[58sniper]

I wrote an article on dealing with back pressure. See it at
Back Pressure: Moving the queue database in Exchange 2007[/quote]

Pat Richard [url=http://mvp.support.microsoft.com/]MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.

http://www.ucblogs.net/blogs/exchange/

 

[hunterdw]

Pat--

I actually read this today. I love the Petri site.

it seems that your article, and most of what I've read, are suggesting the issue is related to "free space"

as in... move these things because your drive is filling up.

In my case, I have plenty of drive size... about 700 gigs free..

All of my storage is virtualized... so, while I could create another datastore, and may still do that, it shouldn't matter. I'm still reading from/writing to the same spindles.

Any other ideas? I'm thinking about disabling resource monitoring on the hub transport... I know that's horrible, but it's annoying

thanks for your quick response...

--DW

 

[hunterdw]

I left a detail off.. sorry...

the attribute causing the issue is "version buckets"

As in... version buckets increase which causes pressure to go from normal to medium and medium to high

when it goes back down to normal, it's cause "version buckets" get closer to normal

Does that help?

 

[hunterdw]

Update: I think I found a good place to start...

http://msexchangeteam.com/archive/2008/05/14/448890.aspx

I adjusted

a) DatabaseMaxCacheSize

We'll see...

--DW

 

[58sniper]

Let us know how it goes. Here's the corrected link from my earlier post: Back Pressure: Moving the queue database in Exchange 2007. The URL was correct, but the formatting of the link was mangled.

Pat Richard MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.

http://www.ucblogs.net/blogs/exchange/

 

[hunterdw]

so far, things are much improved since I adjusting that setting. Even through some intensive testing through backups, mailbox moves, etc. etc.

--DW

 

[58sniper]

Glad to see things are working. If the queue database is on C:, you should follow the article I listed and still move it.

Pat Richard MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.

http://www.ucblogs.net/blogs/exchange/

 

[hunterdw]

Pat,

nothing on C... everything.. DBs, Logs, mailstores, etc. are all on iSCSI mounted storage

I recently did this too...

netsh int ip set chimney DISABLED as refefenced on the EHLO blog

so far, performance is back to "normal"

Next for me is a) jumbo frames on the iSCSI storage and b) getting the last of my Exchange 2003 mailboxes on Exchange 2007 so I can be in a full Exchange 2007 infrastructure.

--DW