Copying seems sluggish at times

[Jesse Delk]

I have a rp7410, VA7110 with 15 36.4 Gb drives,
Autoraid, 3 luns u01, u02, u03

Before issue, copying a certain file (2GB) from u01 to u03 took about a minute. Starting a few days ago (copying same file same size) it is now taking about ten minutes.

I looked at Command View and the status shows "Disk scrubbing in Progress" and "Optimization in Progress".

I did not start these, is this something it starts by itself?

Is this causing the copying to be slower?

What things could I look at, that could be causing the problem?

What error logs could I look at?

[Steven E. Protter · Accepted solution]

Shalom,

/var/adm/syslog/syslog.log

Check the disk array for error messages and errors.

http://www.hpux.ws/system.perf.sh

Need better performance data to lock this down.

SEP

[Steve Steel]

Hi

Disk Scrubbing in Progress

This message indicates that the disk scrubbing is in process. Disk scrubbing is a background task that reads all sectors on all installed disks (it reads the sectors even if they are allocated to LUNs). This forces the disk mechanism to re-map bad tracks, or to fail the read if a bad spot is encountered. If a bad disk is found, it is auto-failed (provided there is redundancy). This will start a rebuild (assuming enough spare space and auto-rebuild is on).


If enabled it will run every five days, completion dependent on host activity

NOTE: Disk scrubbing will take about two hours for 15-18GB disks

Enable or disable the Scrub

armmgr -p {start|disable}

The following shows disabling scrubbing on the array alias Array1.

armmgr -p disable Array1



Since you see this permanently you need indeed to dig deeper for potential HW problems

Always running it would slow you down

Steve Steel

[Jesse Delk]

Syslog gave me some ugly errors. Can someone help me in solving them?

These are the errors when this problem first happened.

See attached


little sample of syslog.log...

Jan 2 02:54:00 orasvr2 su: + tty?? root-oracle
Jan 2 03:13:00 orasvr2 su: + tty?? root-oracle
Jan 2 03:13:07 orasvr2 above message repeats 19 times
Jan 2 03:14:00 orasvr2 su: + tty?? root-oracle
Jan 2 03:15:20 orasvr2 vmunix:
Jan 2 03:15:20 orasvr2 vmunix: SCSI: Read error -- dev: b 31 0x040300, errno: 126, resid: 1024,
Jan 2 03:15:20 orasvr2 vmunix: blkno: 8, sectno: 16, offset: 8192, bcount: 1024.
Jan 2 03:15:00 orasvr2 su: + tty?? root-oracle
Jan 2 03:15:20 orasvr2 vmunix: LVM: Performed a switch for Lun ID = 0 (pv = 0x0000000048506000), from raw device 0x1f040300 (with priority: 0, and current flag
s: 0x40) to raw device 0x1f060300 (with priority: 1, and current flags: 0x0).
Jan 2 03:15:20 orasvr2 vmunix: LVM: VG 64 0x020000: PVLink 31 0x040300 Failed! The PV is still accessible.
Jan 2 03:15:25 orasvr2 vmunix: LVM: VG 64 0x020000: PVLink 31 0x040300 Recovered.
Jan 2 03:15:25 orasvr2 vmunix: LVM: Performed a switch for Lun ID = 0 (pv = 0x0000000048506000), from raw device 0x1f060300 (with priority: 1, and current flag
s: 0x0) to raw device 0x1f040300 (with priority: 0, and current flags: 0x0).

[Andrew Young_2]

Hi Jesse

Are you connecting to your VA7110 with Fibre or copper cables. This behaviour is often an indication that there is either a cable or controller problem. I have seen similar behaviour resulting from a faulty laser diode on a FC controller.

If the link between the VA and the rp7410 has an error the VA will scrub to make sure there is no corruption. If this happenning continually it will seriously impact performance.

First check would be to check all the cables. Thereafter you may need to look at either the VA or the IO card in the rp7410.

You can also view the controller logs on the VA with armlog -e .

HTH

Andrew Y

[Jesse Delk]

Am I doing something wrong here? I'm logged in as root.

# armlog -e VA7110

Login failed, you are not authorized to access this system.

Cables are plugged in. I seemed to have two cards on the rp7410 they are plugged into. Some 2G Fibre channel thing. They both have two green lights on them....and on one of them..one of the green lights is blinking...mean anything?

[Roland Piette]

Hi Jesse,

armlog with '-e' option needs an Endatetimevalue (see man page).

You could also try an armdiscover before to identify the array-id.

Roland

[Jesse Delk]

Thanks for the help. Having HP help me search for a hardware issue.