I'm doing this work right now for a big project here. My first endeavor into
Linux.
The lin_tape drivers for 6.2 will require a .rules file in /etc/udev/rules.d
(or wherever your udev stuff lives) mainly because of the instance owner/group
requirements to run 6.2 dsmserv processes. Unless you can alter your default
udev rules for EVERYTHING, then you'll need the .rule file to assign ownership
and mode parameters for the tape devices.
Mine, so far, looks like this:
#cat /etc/udev/rules.d/98-lin_tape.rules
KERNEL=="IBMchanger*", SYSFS{primary_path}=="Primary",
SYSFS{serial_num}=="0000078150090402", OWNER="tsminst1", MODE="0600",
SYMLINK="lin_tape/IBMchanger137B"
KERNEL=="IBMchanger*", SYSFS{primary_path}=="Alternate",
SYSFS{serial_num}=="0000078150090402", OWNER="tsminst1", MODE="0600",
SYMLINK="lin_tape/IBMchanger138A"
KERNEL=="IBMtape*[0-9]", SYSFS{ww_port_name}=="0x5005076300549127",
OWNER="tsminst1", MODE="0600", SYMLINK="lin_tape/IBMtape137"
KERNEL=="IBMtape*[0-9]", SYSFS{ww_port_name}=="0x5005076300549128",
OWNER="tsminst1", MODE="0600", SYMLINK="lin_tape/IBMtape138"
KERNEL=="IBMtape*[0-9]", SYSFS{ww_port_name}=="0x5005076300549129",
OWNER="tsminst1", MODE="0600", SYMLINK="lin_tape/IBMtape139"
KERNEL=="IBMtape*n",
OWNER="tsminst1", MODE="0600"
There are a lot of gotchas with this method that I'm running into. I'm not
sure if they are kernel bugs or driver issues but not much of this is
documented anywhere. Bullet-list (so far):
- If you have alternate pathing or data path failover, lin_taped needs be
installed and running. Problem is getting persistent binding to work with
this. There's a race condition where once modprobe lin_tape is run, the udev
files are created with the rules. But the SYSFS{primary_path} key isn't
defined correctly until lin_taped is run, BUT lin_taped can't run until
lin_tape is loaded. So by the time lin_taped is executed and running, the
lin_tape rules have already been processed for udev.
o My workaround will be to create an init script that will run lin_taped and
then udevtrigger. Seems to work, but udevtrigger once crashed the system.
- Sometimes when lin_tape is loaded, the mode is incorrect for devices. The
fix is again "udevtrigger".
- KERNEL=="IBMtape*" doesn't work for renaming, because sometimes a symlink
to IBMtape1n is used instead of IBMtape1. Which is why I have the character
class "IBMtape*[0-9]"
Here's the output for ls /dev/ commands for when I believe things are
configured correctly. Caveat: I haven't even tested reading/writing to these
devices yet, let alone defining the devices to TSM.
#ls -l /dev/IBMtape*
crw-r--r-- 1 root root 250, 3071 Sep 27 11:43 /dev/IBMtape
crw------- 1 tsminst1 root 250, 0 Sep 27 11:43 /dev/IBMtape0
crw------- 1 tsminst1 root 250, 1024 Sep 27 11:43 /dev/IBMtape0n
crw------- 1 tsminst1 root 250, 1 Sep 27 11:43 /dev/IBMtape1
crw------- 1 tsminst1 root 250, 1025 Sep 27 11:43 /dev/IBMtape1n
crw------- 1 tsminst1 root 250, 2 Sep 27 11:43 /dev/IBMtape2
crw------- 1 tsminst1 root 250, 1026 Sep 27 11:43 /dev/IBMtape2n
#ls -l /dev/lin_tape
total 0
lrwxrwxrwx 1 root root 14 Sep 27 11:43 IBMchanger137B -> ../IBMchanger0
lrwxrwxrwx 1 root root 14 Sep 27 11:43 IBMchanger138A -> ../IBMchanger1
lrwxrwxrwx 1 root root 11 Sep 27 11:43 IBMtape137 -> ../IBMtape2
lrwxrwxrwx 1 root root 11 Sep 27 11:43 IBMtape138 -> ../IBMtape0
lrwxrwxrwx 1 root root 11 Sep 27 11:43 IBMtape139 -> ../IBMtape1
HTH,
Sergio
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Zoltan Forray/AC/VCU
Sent: Wednesday, September 22, 2010 2:53 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: [ADSM-L] Looking for SAN/tape experts assistance
I have mentioned in previous posts that we are putting up 2-new RH Linux
based TSM server . These are the first of my existing 5-Linux servers to
use EMC SAN storage.
With every new adventure, we get new problems. This one is driving
everyone crazy and hope someone out there can point us in the right
direction.
We have seen posts in ADSM-L that sorta talk about it, but nothing that
explains what is going on with us or how to resolve it.
Both new servers have been configured identically when it comes to the OS
(RedHat Linux 5.5 kernel 2.6.18-194.11.3.el5) software and other hardware
supporting software (EMC Powerpath and IBM lin_tape drivers - 1.41.1 for
the TS1120/1130 drives)
The problem is this.
Every time we reboot one of the new servers, the values in
/proc/scsi/IBMtape is different in the assignment of /dev numbers to the
drives. It seems to find the tape drives in a different order each time.
None of my 5-production nor the other new TSM server have this problem (I
have rebooted the 2nd new server 4-times and the /dev/IBMtape? values stay
the same).
When looking through the "fixlist" for lin_tape (usually
engineering-speak), we saw this interesting entry at the 1.37 level:
Removed persistent naming script in favor of new method
Questions come to mind about things like "what naming script"......."what
new method" .... "could this possibly be related to what we are
experiencing"?
We have spent all day trying to figure this wrinkle out. Any suggestions
are greatly appreciated.
Zoltan Forray
TSM Software & Hardware Administrator
Virginia Commonwealth University
UCC/Office of Technology Services
zforray AT vcu DOT edu - 804-828-4807
Don't be a phishing victim - VCU and other reputable organizations will
never use email to request that you reply with your password, social
security number or confidential personal information. For more details
visit http://infosecurity.vcu.edu/phishing.html
|
