System Healtchek Report Problems 1

[PSS SYS ADMIN]

Hi everyone,
I've runned the lite performance option of shc on a hpux 11.00 L2000 system.
The result of the configuration area was:
-The LIFS area on LVM mirrored boot disks is not identical.
- /tmp file system is not mounted using the tmplog mount logging option.
- The VxFS parameter vx_ninode is set to the default value of zero.
- LVM Mirror write cache miss rate is too high.
- Multiple swap partitions found on the same device.
- Inactive LVM Volume groups detected.
- Mirrored Logical Volumes have been implemented without Physical Volume Groups.
- One or more boot devices missing from /stand/bootconf file.


Please help me to understand the question and to solve it

Regards...
PSS

[Alex Lavrov.]

I suggest you to read the manuals about each issue. There is just too much information. Run a search in http://docs.hp.com or in google about each issue.
For example "lif area":
http://www.google.com/search?hl=en&lr=&q=lif+area

or "mount tmplog":
http://www.google.com/search?hl=en&q=mount+tmplog&btnG=Google+Search

First links of each search will provide you the info you are looking for.

Good luck.

[Florian Heigl (new acc)]

There's a lot not in 'best shape' on Your system, I'll try to give some notes on all parts, but honestly if this is a production system, You should rather spend the money for the extended analysis by HP, that is what SHC is for, because I don't know everything about the issues. (I'll mark where I'm not sure)


-The LIFS area on LVM mirrored boot disks is not identical.

gather Your boot disks:
lvlnboot -v
Example:
Boot: lvol1 on: /dev/dsk/c0t0d0
/dev/dsk/c1t0d0
(Your disk names will differ)
now do a lifs /dev/rdsk/c0t0d0 and /dev/rdsk/c1td0d0 and compare the results. (There should be no differences, only the file 'PAD' missing is tolerable)


- /tmp file system is not mounted using the tmplog mount logging option.
this is only a performance issue, tmplog means that journaling for this filesystem will only happen in ram. this recommendation is very reasonable, as /tmp only contains volatile data and some software (Gnu autoconf) will create >1000 files/minute there. but noone forces You to change it.


- The VxFS parameter vx_ninode is set to the default value of zero.
this value sets the size of the kernel inode cache. I'm quite unsure here, even with setting "0" there is an upper limit imposed to my knowledge. Definitely a point to ask HP or read docs.

- LVM Mirror write cache miss rate is too high.
I don't know where this value is gathered - really interesting.
Anyway, what they're saying is that Your mirror disks don't keep up with syncing, it takes longer than it should take and for most of the time they appear to be out of sync. this can have fatal consequences on a system crash with outstanding I/O.
look into man lvchange for 'MWC'

- Multiple swap partitions found on the same device.
this could also reduce Your overall system performance, *if* the swap gets used. take care to set the 'pri=N' in /etc/fstab option correctly:
swap space0 gets pri=0
swap space1 gets pri=1
and so forth.
if You can, spread the swap spaces over separate physical devices (but avoid external devices) and then use pri=0 for half of them and so on...

- Inactive LVM Volume groups detected.
unless this is a non-active cluster partner, this is strange. check through the output of vgdisplay and vgdisplay -v, the inactive vg's will be mentioned.
If You try activating them, do this in read-only mode for the beginning, using
vgchange -a r
(if You feel unsure here, let HP do it if they have remote access - You don't want to lose data ;)

- Mirrored Logical Volumes have been implemented without Physical Volume Groups.
This does not have to be a problem.
Using 'PVG's is very adviseable if this is a VG LVM-mirrored over two external arrays or something huge like that. I wouldn't care if this is just two disks next to each other in a disk array.

- One or more boot devices missing from /stand/bootconf file.
this won't interrupt booting or daily work, but can cause serious trouble if Your LVM blows up and You needed to boot from a device that is *not* mentioned in there.

It can be resolved, but might take extra time - down time.
see man bootconf
(when looking for an example I noticed we had it wrong on one host, too.)