From: Donncha O Caoimh (donncha.ocaoimh at domain tradesignals.com)
Date: Fri 14 Jun 2002 - 13:44:12 IST
Over the past 3 days our tape backup server has been acting strangely.
When a backup is in progress to the SCSI tape drive the machine becomes
unresponsive and eventually has to be rebooted. A lot of error messages
appear on the console but not in the syslog (my apologies, I didn't write
them down so this bit is vague!). The errors refer to /dev/hda (the drive on
which everything is installed) being busy and AFAIR similar messages appear
for /dev/hdb (the drive I use to cache Amanda dumps before writing to tape..)
There's something about a status code 80 mentioned too.
This morning, after rebooting the box I ran amflush which basically writes out
everything on /dev/hdb to tape and that worked fine but when I tried to do a
backup afterwards the error messages appeared again.
I suspect it's a hardware conflict somewhere, I'm using a PCI scsi card, PCI
network card (data is read from the network for backups). /proc/interrupts
doesn't show anything strange. I *had* dma enabled on the two IDE drives but
disabled that when the trouble started, here's the hdparm that gets run on
each bootup (same for hdb)
/sbin/hdparm -m16 -d0 -c1 /dev/hda
It all worked fine for several months, although I might have only recently
enabled dma (in the past 4 weeks that is, not immediately before these
problems showed up..)
I'm going to do another backup now and will post more detailed error messages
This archive was generated by hypermail 2.1.6 : Thu 06 Feb 2003 - 13:17:15 GMT