Re: Reducing SCSI timeouts?
If any of the hardware guys are listening, it is a complete bear that the tapeutil functions of ITDT require root access. The old tapeutil didn't. Enterprise shops such as the one I work in are completely paranoid and will not give out root access to non-OS admins except under extremely limited circumstances, and with a good deal of bureaucratic form filling and multiple approvals. Those same shops have also been known to penny-pinch in the past and not purchase the extra-cost Web Server option on libraries. Combine that with a datacentre many hundreds of miles from where the admin is and operators whose expertise amounts to being able to open the CAP on a library and remove the tapes and you have an ironclad case for the old way of doing things. Please, make this change. Thanks Steve Steven Harris TSM Admin, not trusted by the Unix team Canberra Australia On 5/07/2012 7:53 PM, Stout, Susie (NIH/CIT) [E] wrote: Another suggestion: send your commands directly to the hardware. We also have LTO4s (and AIX); I use tapeutil, which comes with the device drivers; you can do most anything with the line commands, including FSF (forward space file), eject a tape, etc.; I often use it to move a cart to/from slot/door ports. Tapeutil is sunsetted and may not be available; IBM now offers ITDT which I am not using -- it apparently comes with some kind of GUI; here's my notes on others' contributions to the List re ITDT, though I've not installed it myself: ITDT notes: to download: Go to http://www.ibm.com/support/fixcentral IBM Fix Central / portal to downloads use these pull-down menus to get to the correct download as follows: 1. In the first pull down menu labeled "Product Group" select "Storage Systems" 2. In the next pull down menu that appears which is labeled "Product Family" select "Tape Systems" 3. With the next pull down menu, "Product Type", select "Tape Device Drivers and Software" 4. This will in turn bring up the "Product" menu 1). This menu allows you to select the "ITDT" 2). Platform: select your operating system (generic: ie. Linux) and will show all device drivers for that platform 3). The next screen allows narrowing the search if needed; click "Continue" to see what's available Supposedly tapeutil isn't gone. It's not "replaced" by itdt either. itdt contains the tapeutil command: run "itdt -f /dev/smc0 inventory", or run itdt and choose "U" to get into the text-gui version. Caveat: itdt is neither installed, nor does it like being run from anywhere other than its own directory...put this in your .profile and you'll be set: tapeutil() { (cd /opt/IBMtape ; ./itdt $* ;) ; } (assumes you unpacked ITDT into /opt/IBMtape. Adjust as necessary for your environment); the function is because you're passing parameters to it; subshell because we don't want to stay in the dir where it's installed. Hope you find the notes useful, Alex; it's possible there may be a SCSI timeout setting as well...and many thanks to others for the ITDT contributions above! - Susie -- Date:Mon, 2 Jul 2012 08:50:43 -0700 From:Alex Paschal mailto:apasch...@frontier.com>> Subject: Re: Reducing SCSI timeouts? Hi, Sascha. I don't know how to reduce the timeout, but IPLing the drives = should allow the process to cancel. I used to just walk back into the data= center and flip the power switch, it got me up from my desk, but you can IP= L the drive from the 3584 Web Specialist also. On 7/2/2012 7:26 AM, Sascha Askani wrote: Hi everybody, due to faulty tape drives, we have "lost" some of our LTO4 tape=20 volumes, i.e. these volumes are no longer readable by any of our other=20 tape drives. AUDIT VOLUME sometimes stalls at the first file to check=20 and throws errors and gets reduced to a crawl. The only way out for us is currently "DELETE VOLUME DISCARDD=3DYES", but= =20 for obvious reasons, I do a "MOVE DATA" beforehand to rescue the data=20 that any of the drives is able to read. (No copy pools were yet=20 configured) Since we have "a couple of" volumes to check, I'd like to get some=20 insight if it is possible to reduce the SCSI timeout value for the=20 tape drives, since we have to wait ~15 minutes for a single "CANCEL PROCE= SS" I tried: echo "180" > /sys/class/lin_tape/IBMtapeX/device/timeout But that didn't work out. Any hints? * TSM for Linux/x86_64 - Version 6, Release 2, Level 3.100 * RedHat 5 * lin_taped 1.68.0-1 Thanks in advance! BR, Sascha
Reducing SCSI timeouts?
Another suggestion: send your commands directly to the hardware. We also have LTO4s (and AIX); I use tapeutil, which comes with the device drivers; you can do most anything with the line commands, including FSF (forward space file), eject a tape, etc.; I often use it to move a cart to/from slot/door ports. Tapeutil is sunsetted and may not be available; IBM now offers ITDT which I am not using -- it apparently comes with some kind of GUI; here's my notes on others' contributions to the List re ITDT, though I've not installed it myself: ITDT notes: to download: Go to http://www.ibm.com/support/fixcentral IBM Fix Central / portal to downloads use these pull-down menus to get to the correct download as follows: 1. In the first pull down menu labeled "Product Group" select "Storage Systems" 2. In the next pull down menu that appears which is labeled "Product Family" select "Tape Systems" 3. With the next pull down menu, "Product Type", select "Tape Device Drivers and Software" 4. This will in turn bring up the "Product" menu 1). This menu allows you to select the "ITDT" 2). Platform: select your operating system (generic: ie. Linux) and will show all device drivers for that platform 3). The next screen allows narrowing the search if needed; click "Continue" to see what's available Supposedly tapeutil isn't gone. It's not "replaced" by itdt either. itdt contains the tapeutil command: run "itdt -f /dev/smc0 inventory", or run itdt and choose "U" to get into the text-gui version. Caveat: itdt is neither installed, nor does it like being run from anywhere other than its own directory...put this in your .profile and you'll be set: tapeutil() { (cd /opt/IBMtape ; ./itdt $* ;) ; } (assumes you unpacked ITDT into /opt/IBMtape. Adjust as necessary for your environment); the function is because you're passing parameters to it; subshell because we don't want to stay in the dir where it's installed. Hope you find the notes useful, Alex; it's possible there may be a SCSI timeout setting as well...and many thanks to others for the ITDT contributions above! - Susie -- Date:Mon, 2 Jul 2012 08:50:43 -0700 From:Alex Paschal mailto:apasch...@frontier.com>> Subject: Re: Reducing SCSI timeouts? Hi, Sascha. I don't know how to reduce the timeout, but IPLing the drives = should allow the process to cancel. I used to just walk back into the data= center and flip the power switch, it got me up from my desk, but you can IP= L the drive from the 3584 Web Specialist also. On 7/2/2012 7:26 AM, Sascha Askani wrote: > Hi everybody, > > due to faulty tape drives, we have "lost" some of our LTO4 tape=20 > volumes, i.e. these volumes are no longer readable by any of our > other=20 tape drives. AUDIT VOLUME sometimes stalls at the first file > to check=20 and throws errors and gets reduced to a crawl. > > The only way out for us is currently "DELETE VOLUME DISCARDD=3DYES", > but= =20 > for obvious reasons, I do a "MOVE DATA" beforehand to rescue the > data=20 that any of the drives is able to read. (No copy pools were > yet=20 > configured) > > Since we have "a couple of" volumes to check, I'd like to get some=20 > insight if it is possible to reduce the SCSI timeout value for the=20 > tape drives, since we have to wait ~15 minutes for a single "CANCEL > PROCE= SS" > > I tried: > > echo "180" > /sys/class/lin_tape/IBMtapeX/device/timeout > > But that didn't work out. Any hints? > > * TSM for Linux/x86_64 - Version 6, Release 2, Level 3.100 > * RedHat 5 > * lin_taped 1.68.0-1 > > Thanks in advance! > > > BR, > > Sascha >
Re: Reducing SCSI timeouts?
Hi, Sascha. I don't know how to reduce the timeout, but IPLing the drives should allow the process to cancel. I used to just walk back into the datacenter and flip the power switch, it got me up from my desk, but you can IPL the drive from the 3584 Web Specialist also. On 7/2/2012 7:26 AM, Sascha Askani wrote: Hi everybody, due to faulty tape drives, we have "lost" some of our LTO4 tape volumes, i.e. these volumes are no longer readable by any of our other tape drives. AUDIT VOLUME sometimes stalls at the first file to check and throws errors and gets reduced to a crawl. The only way out for us is currently "DELETE VOLUME DISCARDD=YES", but for obvious reasons, I do a "MOVE DATA" beforehand to rescue the data that any of the drives is able to read. (No copy pools were yet configured) Since we have "a couple of" volumes to check, I'd like to get some insight if it is possible to reduce the SCSI timeout value for the tape drives, since we have to wait ~15 minutes for a single "CANCEL PROCESS" I tried: echo "180" > /sys/class/lin_tape/IBMtapeX/device/timeout But that didn't work out. Any hints? * TSM for Linux/x86_64 - Version 6, Release 2, Level 3.100 * RedHat 5 * lin_taped 1.68.0-1 Thanks in advance! BR, Sascha
Reducing SCSI timeouts?
Hi everybody, due to faulty tape drives, we have "lost" some of our LTO4 tape volumes, i.e. these volumes are no longer readable by any of our other tape drives. AUDIT VOLUME sometimes stalls at the first file to check and throws errors and gets reduced to a crawl. The only way out for us is currently "DELETE VOLUME DISCARDD=YES", but for obvious reasons, I do a "MOVE DATA" beforehand to rescue the data that any of the drives is able to read. (No copy pools were yet configured) Since we have "a couple of" volumes to check, I'd like to get some insight if it is possible to reduce the SCSI timeout value for the tape drives, since we have to wait ~15 minutes for a single "CANCEL PROCESS" I tried: echo "180" > /sys/class/lin_tape/IBMtapeX/device/timeout But that didn't work out. Any hints? * TSM for Linux/x86_64 - Version 6, Release 2, Level 3.100 * RedHat 5 * lin_taped 1.68.0-1 Thanks in advance! BR, Sascha