Re: LVM2 Boottime Performance Issue

2007-07-11 Thread Brad Hinson
On Wed, 2007-07-11 at 12:43 +, Rick Truett wrote: > I have gotten a number of performance monitor questions. > > I will get some VM Monitor information on channels and disk performance and > feed this back. > > How can I discover where in lvscan I am spending my time? > One idea is to remove

Re: LVM2 Boottime Performance Issue

2007-07-11 Thread Rick Truett
I have gotten a number of performance monitor questions. I will get some VM Monitor information on channels and disk performance and feed this back. How can I discover where in lvscan I am spending my time? What I can answer at the moment is; The Linux guest has 3 IFLs on a z9 processor wi

Re: LVM2 Boottime Performance Issue

2007-07-10 Thread Rick Troth
The [E]CKD model does not lend itself to this. I am not completely sure why (because the overhead of applying and then peeling off track and record boundaries doesn't seem to be THAT intrusive to cause so much pain). Whether [E]CKD or FBA (eg: SAN), the main point is that you need larger volumes

Re: LVM2 Boottime Performance Issue

2007-07-10 Thread Rob van der Heij
On 7/10/07, Brad Hinson <[EMAIL PROTECTED]> wrote: Using mod-9 instead of mod-3 gives you a performance hit, especially if they're not spread across multiple I/O channels. Check out the latter sections of this PDF for more info: I doubt that's the issue here. He's talking about 130 times -9 a

Re: LVM2 Boottime Performance Issue

2007-07-10 Thread Brad Hinson
Using mod-9 instead of mod-3 gives you a performance hit, especially if they're not spread across multiple I/O channels. Check out the latter sections of this PDF for more info: http://linuxvm.org/Present/SHARE102/s9300jo.pdf It's old, but the info is still relevant. In addition to spreading ou

Re: LVM2 Boottime Performance Issue

2007-07-10 Thread Dave Jones
Hi, Rick. Can you provide any more detailed information about this problem? You already know that it appears to be in the lvscan process, but can we find out *where* in the lvscan processing the time is being spent? Is the Linux virtual machine getting enough resources at boot time? How loaded ar

LVM2 Boottime Performance Issue

2007-07-10 Thread Rick Truett
Hello; Here is the environment: z/VM 5.2, RHEL 4, LVM2 with fullpack 3390-9 mini-disks EXT3 filesystems. There are 4 lvgroups each with ~130 volumes (give or take a few). Issue: On boot of the Linux system it takes upward to 80 minutes to open the LVM filesystems. Most of the time is in lvscan