Hey Karl, Did you ever mention in this thread how much RAM?
How much SWAP / PAGING space on disk, and on which disks the swap is defined? Where is the /tmp defined in uvconfig? What kind of file system is /tmp, and how big? When was the last time you rebuilt that file system? Or the filesystem that contains these files? Is this the same file system where the rest of the UV Accounts data files reside, and where the standard hashed files resided before you converted to Dynamic file types? When was the last time this file system was saved to tape, reformated, and restored from tape afterwards? What kind of file system is /dev/lv00? Are1024-blocks a disk I/O frame size, or the standard du utility reporting blocks in 1024 size chunks? What is the io file system frame size and any interleave factors? Any RAID involved? I think you mentioned two disks, (i.e. "both disk loads"), Can you please share how many physical disks and their sizes, for everything on this system, OS, and UV Database? And their controllers? How about your uvconfig? AIX 4.3.3.0 with no patches? <scratching the goatee.... and squinting a suspect eye...> You may want to load up the DPMonitor for it's 10 day Free Trial Evaluation, system-wide license, (Performance Agent for the AIX Server, and Perfomance Explorer for a Windows PC workstation connected to the same network as the Server). Obviously, activity on this platform has pushed through some threshold level, and has your performance in a tizzy... %<---- below from previous posts on this thread: * * * * * * * * * * Platform: IBM 7043-140, AIX 4.3.3.0, UV 10.0.14 with 42 licenses. df -k: Filesystem 1024-blocks Free %Used Iused %Iused Mounted on /dev/lv00 3072000 1168752 62% 14935 2% /u1 I've done no patching to this OS and the above file system is the standard IBM journaled type, I believe. * * * * * * * * * * ----- Original Message ----- From: "Karl L Pearson" <[EMAIL PROTECTED]> To: "u2-users" <[EMAIL PROTECTED]> Sent: Wednesday, July 21, 2004 12:14 PM Subject: Re: [U2] (UV) Analyze.shm -s (Unclassified) > On Wed, 2004-07-21 at 03:24, Martin Phillips wrote: > > > We may also be looking for the wrong thing here. A semaphore collision > > count of 6 really is nothing to get excited over, especially if the system > > had been up for a while. Looking back to Karl's original posting, I think > > we need to know a bit more about how the "performance hits" exhibit > > themsleves. > > I've done a bit of checking and came to the conclusion that IO was NOT > an issue, as both disks loads are very even and in the low teens to low > 20%-ile of use. > > I changed the uniVerse tunable FSEMNUM from 23 to 53. Today, so far, the > load average is in the low 2s instead of in the 3s and 4s. Of course we > only have 30 of 42 users, so time will tell... > > The symptoms of the performance problem is that users notice things > going very slowly, especially selects and sorts, but also when typing > text there are pauses of parts of a second to a couple seconds when the > system is very sluggish. > > -- > Karl L. Pearson > Director of IT, > ATS Industrial Supply > Direct: 801-978-4429 > Toll-free: 888-972-3182 x29 > Fax: 801-972-3888 > http://www.atsindustrial.com > [EMAIL PROTECTED] > ------- > u2-users mailing list > [EMAIL PROTECTED] > To unsubscribe please visit http://listserver.u2ug.org/ ------- u2-users mailing list [EMAIL PROTECTED] To unsubscribe please visit http://listserver.u2ug.org/
