August 8, 2018 at 3:17 am
The SQL instance RPT01 is consistently under internal memory pressure on 1 of the 2 Numa nodes. This is best represented by the PLE perfmon counters:
PLE on Numa Node 0 resets to 0 every 60secs
- Intermittently this causes “A significant part of sql server process memory has been paged out. This may result in a performance degradation….”
PLE on Numa Node 1 keeps growing as expected
- No problem
The Cluster Configuration
· 2 Server Node Windows Server 2012 Failover Cluster
· 9 SQL clustered instances (ranging from SQL 2008 R2 - 2016)
· Each Server Node has 512GB of RAM (typically 300GB always available - every instance has Max Memory constrained)
· 2 Numa Nodes with 12 Core / node with HT enabled (total of 48CPUs)
Problem SQL Instance (RPT01)
· SQL Server 2012 Enterprise Edition (11.0.3000.0)
· Min Memory 0, Max Memory 8GB
· Lock Pages in Memory (LPiM)
· 3 user databases (total size less than 100MB)
· Very low usage
The following changes attempted in troubleshooting(without success)
1. Offlined user databases, stopped SQL Agent (i.e. stopped all known connections to instance)
2. Switched to other node and back again
3. Removed LPiM
4. Increased Max Memory from 8GB to 10GB
5. Cleared Buffer Pool & Cache repeatedly (DROPCLEANBUFFERS & FREEPROCCACHE)
August 14, 2018 at 7:41 am
If the problem node is part of a cluster, what expected role would it play in the clustering, and/or failover process? Seems to me a 10 GB memory node wouldn't be capable of much in that regard, so maybe it shouldn't be participating at such a seriously limited capacity ? I just suspect you may be asking too much of a node that small given what the rest of your instances might look like... and perhaps solely based on what it could be expected to do in a failover situation....
Steve (aka sgmunson) 🙂 🙂 🙂
Rent Servers for Income (picks and shovels strategy)
Viewing 2 posts - 1 through 1 (of 1 total)
You must be logged in to reply to this topic. Login to reply