Low physical memory alert on 3 servers - erroneous ..?

fatherjack2

Hi,

I have 3 SQL 2005 Server Instances that are monitored by SQL Response. Server Instances A + B are on one physical box and server instance C is on its own box and hosts the SQL Response repository too.

All three instances reported a Low Physical memory alert at the same time (within 5s) this morning at the time that I changed the SQL Response Alert Repository Service login account (on installation I had used my credentials and I need to change my pwd so moved the service to another - more appropriate - domain account.

I dont think that we would have got close to using all the RAM on these servers, servers A + B share 20GB and server C has 8GB.

Is this something that needs investigating or a known outcome when the service account is changed?

cheers

Jonathan

dlkj

Hi FatherJack,

What does the graph for the performance snapshot in the alert details show?

Cheers,
--
Daniel

fatherjack2

Daniel,

the pie chart that shows SQL Server 18.4 GB, other processes 1.4GB and other 192MB on servers A + B (they are set to have a maximum of 9GB each, thus leaving 2GB for OS etc) at 10:14:24

and for server C ...
at 10:14:19 SQL Server 6.3GB, other processes 1.5GB, free 235.1MB

regards

Jonathan

dlkj

The system processes section should provide a breakdown of windows process memory usage. Does anything there look unusual?

fatherjack2

nothing really stands out;

I would sort it by the mem usage or VM size columns if I could so as to see the worst offenders ... feature hint

I would send it to you if I could select from that pane and paste into Excel ... feature hint

I dont know what would be normal for any of the 70+ processes on the servers though .

background:
Servers A and B host OLTP and OLAP activities for our main company system (maybe max 200 concurrent users via 2 IIS servers),

Server C supports all other database services for the company - 45 databases, max 10-20 concurrent users in maybe 3 or 4 of these systems - HR, Accounts, Intranet, etc., )

certainly both server (h/w) are doing very different things and are (at this time of day) totally unconnected. Over night there are data transfer scripts that run.