Application server directory brief hang

[Thomas Maddox]

Can anyone help with this?

A directory named /uo1/app/tomax/data hung for about 10 minutes. Could not do commands associated with that directory such as ls or bdf. Syslog had several of the following messages:

"Mar 11 14:09:10 ebb vmunix: LVM: Performed a switch for Lun ID = 0 (pv = 0x00000
00043e3f080), from raw device 0x1f0bc500 (with priority: 0, and current flags: 0
x40) to raw device 0x1f0ac500 (with priority: 1, and current flags: 0x0).
Mar 11 14:09:10 ebb vmunix:
Mar 11 14:09:10 ebb vmunix: SCSI: Read error -- dev: b 31 0x0bc500, errno: 126,
resid: 2048,
Mar 11 14:09:10 ebb vmunix: blkno: 8, sectno: 16, offset: 8192, bcount: 2048
.
Mar 11 14:09:12 ebb inetd[28494]: registrar/tcp: Connection from ebb (10.0.0.62)
at Tue Mar 11 14:09:12 2003
Mar 11 14:09:13 ebb vmunix: LVM: vg[1]: pvnum=0 (dev_t=0x1f0ac500) is POWERFAILE
D
Mar 11 14:09:18 ebb vmunix: LVM: Recovered Path (device 0x1f0bc500) to PV 0 in V
G 1.
Mar 11 14:09:18 ebb vmunix: LVM: Performed a switch for Lun ID = 0 (pv = 0x00000
00043e3f080), from raw device 0x1f0ac500 (with priority: 1, and current flags: 0
xc0) to raw device 0x1f0bc500 (with priority: 0, and current flags: 0x80).
Mar 11 14:09:18 ebb vmunix: LVM: Recovered Path (device 0x1f0ac500) to PV 0 in V
G 1.
Mar 11 14:09:13 ebb vmunix: LVM: vg[1]: pvnum=0 (dev_t=0x1f0ac500) is POWERFAILED"

The server is used as an oracle app server.
A bdf -i for this file system shows.
/dev/vg01/lvol2 10240000 2707718 7296924 27% 956 941532 0% /uo1/app/tomax

Cannot tell what was running at that time.

This runs on a VA7400, which has been referred to as "An overactive array" by HP. I ran a getVAlogs script and I am interested in passing on that data or any other data you need.

[Ross Zubritski]

WHOA........

Power Failed is a bad thing my friend. Get HP out to replace that defective device.

Regards,

RZ

[Steven E. Protter]

Looks like you had a powerfail on a drive and your system failed over via PVLINKS.

run dmesg or

mstm
stm
xstm(In X windows only)

Exercize every device and card and see if you find something dead.

Take a good make_tape_recovery Ignite backup, and cold backup of your database. FAST!

SEP

[Thomas Maddox]

The array has run successfully for 1 week without a repeat of the error. I have contacted HP and we are looking at the results of the getVAlogs.txt script. No errors show in stm. No repeat mesages in syslog. Disk timeouts on the fibre channels are set to the default on for this system. Does anyone have anything new and interesting to add?