Hi, I had some concerns about a possible memory leak on a 220R (see below) Many thanks for the reassuring responses I received, all of which indicated that my machine was behaving in a normal fashion (for a pre Solaris 8 box, that is). Pyers Simon quoted Cockroft & Petit's "Sun Performance and Tuning": - "When you look at vmstat, please don't waste time worrying where all the RAM has gone. After a while the free list will stabilise at around 3% (three) [which in your case is around 60MB] of the total memory configured. The system stops bothering to reclaim memory above this level, even when you aren't recovering anything" The buffer cache will consume free memory (largely for file data caching) and applications requiring memory will reclaim it as required. The main indicators of truly strained memory resources are excessive page scanning/swapping rates. Kevin Buterbaugh indicated that so long as the scan rate didn't exceed 273 pages per second for extended periods then the system was healthy. Michael Hocke suggested running Virtual Adrian or vmstat over time to monitor these stats. Darren Dunham suggested possibly turning on priority paging in the kernel. Pyers Simon also sent me a copy of "The Solaris Memory System" by Richard McDougall, which is currently being read ... : - http://www.sun.com/sun-on-net/performance/vmsizing.pdf My thanks also to John Riddoch, Graham Rauxon and Peter Ondruska for their speedy responses. -Stephen. -=-=-=-=Original Question=-=-=-=-=-=-=-= Hi all, I have a 220R running 2.6 105181-28. It has 2gb of RAM. It is running as our internal DNS master (BIND 8-22p5), NIS master and mail server. We are running qmail 1.03 and UW imap 2001a. The problem is that the physical RAM is constantly close to being maxed out - it usually sits at about 32mb free. The biggest hog appears to be imap (at present, 162 client sessions running, averaging about 5mb usage), but even on slow days with far fewer imap clients the free RAM count is still down around 32mb. I've seen references to file system caching being a possible cause of this, but the implication is that this usually happens over time - this machine maxes out within about 20 minutes of a reboot. The automounter is mounting all the imap users' home directories, so I'm wondering if it's an issue with nfs file caching. I've run SymbEL zoom on the box and it doesn't squawk about the ram consumption. vmstat over time reports healthy scan rates. I suppose my question is: - Is this normal behaviour or am I looking at a memory leak? If so, what are the optimal tools to detect the cause? I've done some analysis using kmastat/crash, but nothing is leaping out at me - I've been running comparisons with other, healthy boxes and the bucket usage patterns appear consistent. As this is a production machine, my actions in trying to isolate the fault have been restricted - should I be thinking of an out of hours reboot and then bringing up the services incrementally while monitoring the mem stats? TIA for any help/info. Stephen Ryan IT Group Parthus Technologies plc ph: +353-1-402 2579 fax: +353-1-402 5711 _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Wed Apr 3 05:42:20 2002
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:39 EST