Trouble with tape changer
Hi all. If you think you might be able to help me get Amanda working with a Sun server and tapechanger, please read on and let me know what you think. I inhereted a project to configure Amanda to back up a Sun E6500 with an attached StorEdge 30-slot changer, which includes two Quantum DLT drives. Until last Thursday, I had very little experience with Amanda, so bear with me if I missed someting obvious. Using the archives for this group, I have been unable to get around a couple of roadblocks. I can use mtx successfully to work with the drive: gorge# mtx -f /dev/changer inquiry Product Type: Medium Changer Vendor ID: 'ATL ' Product ID: 'P10006220070' Revision: '2.01' Attached Changer: No (mtx status lists the two drives and thirty slots, with labels and status of tapes) ... but I can not get the amcheck or amtape programs to work correctly: gorge# amcheck -s Test Amanda Tape Server Host Check - Holding disk /public/amanda: 16282266 KB disk space available, that's plenty amcheck-server: could not get changer info: badly formed result from changer: /dev/rmt/1 (brought to you by Amanda 2.4.3b2) gorge# amtape Test current amtape: could not get changer info: badly formed result from changer: /dev/rmt/1 Yet, mt can communicate with the drives ... gorge# mt -f /dev/rmt/1 status Quantum DLT7000 tape drive: sense key(0x0)= No Additional Sense residual= 0 retries= 0 file no= 0 block no= 0 (same result for /dev/rmt/2) ... and the chg-zd-mtx script seems to work ... gorge# ./chg-zd-mtx -info 2 30 1 1 Do you have any ideas as to why I am getting the badly formed errors above? I feel like I'm close to getting this system up and running, but I keep banging my head on different issues, and this is the latest one. I have included relevant portions of the configs below. Thanks very much in advance, Eric amanda.conf: runtapes 1 tpchanger /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Test/chg-zd-mtx tapedev /dev/rmt/1 changerfile /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Test/changer changerdev /dev/changer changer: # this is the changer file # multieject 0 gravity 0 needeject 1 ejectdelay 30 statefile changer-status firstslot 1 lastslot 30 slot 1 /dev/rmt/1 slot 2 /dev/rmt/2
Re: a tape to do experiments with, but not in the tapecycle
On Tue, 26 Feb 2002, Tom Van de Wiele wrote: My question is this: how would I add the tape to the list, so I can use if freely when I want, AND (more important) NOT disturb the tapecycle I have now. Come to think of it, it doesn't need to be in the tapecycle, because I only want to use it when I want to. Any ideas? I would simply copy over your current config directory to a new one, rename it to something appropriate, and run your tests with a whole different configuration. You can leave things largely the same, and you have the ability to label another tape without disrupting your current batch. - Eric Trager
Multi-tape question
Hi, all. I managed to get my changer working with Amanda using chg-scsi. Today I started the first test dump to see what would happen. Using hardware compression, my DLT-IV tapes hold about 70 GB. I set up a dump of several large and small partitions totalling ~88 GB. One was not mounted for the dump so the total was actually 73 GB. I set runtapes to 2 in the amanda.conf file thinking this would tell amanda to use two tapes, but I see this in the amdump log: [snip] INITIAL SCHEDULE (size 73008621): gorge /export/home pri 11747 lev 0 size 16340165 gorge /usr/ pri 11747 lev 0 size 15425329 gorge /usr/oracle2 pri 11747 lev 0 size 12874971 gorge /usr/oracle pri 11747 lev 0 size 7166160 gorge /usr/oracle5 pri 11747 lev 0 size 4502198 gorge /usr/oracle3 pri 11747 lev 0 size 4348242 gorge /usr/oracle4 pri 11747 lev 0 size 4143069 gorge / pri 11747 lev 0 size 3541838 gorge /public pri 11747 lev 0 size 982111 gorge /var pri 11747 lev 0 size 818101 gorge /export/gorge pri 11746 lev 0 size 2865347 gorge /export/mada/yoga pri 11746 lev 0 size 40 DELAYING DUMPS IF NEEDED, total_size 73008621, tape length 69029888 mark 43 planner: FAILED gorge /export/gorge 20020226 0 [dumps too big, but cannot incremental dump new disk] planner: FAILED gorge /var 20020226 0 [dumps too big, but cannot incremental dump new disk] planner: FAILED gorge /public 20020226 0 [dumps too big, but cannot incremental dump new disk] delay: Total size now 68342837. [snip] So it dropped some small partitions until it got to a total it likes. I thought, however, that runtapes tells Amanda to use the changer to fill additional tapes as needed. Am I assuming the wrong thing? Does runtapes mean something else? I have included some snips from my configs. Please shed some light on this if you can. TIA, Eric amanda.conf runtapes 2 # number of tapes to use in a single run of amdump tpchanger /usr//opt/amanda-2.4.3b2/etc/amanda/Othertest/chg-scsi tapedev 0 changerfile /usr//opt/amanda-2.4.3b2/etc/amanda/Othertest/chg-scsi-solaris.conf labelstr ^CMAG-[0-9][0-9]*$ dumpcycle 1 #the number of days in the normal dump cycle bumpdays 1 #minimum days at each level bumpsize 20 Mb #minimum savings (threshold) to bump level 1 - 2 bumpmult 2 #threshold = bumpsize * bumpmult^(level-1) runspercycle 1 # the number of amdump runs in dumpcycle days tapecycle 4 # the number of tapes in rotation chg-scsi-solaris.conf number_configs 1 eject 1 # Tapedrives need an eject command sleep 45 # Seconds to wait until the tape gets ready cleanmax1000# How many times could a cleaning tape get used changerdev /dev/changer # config 0 drivenum0 dev /dev/rmt/1cbn startuse0 # The slots associated with the drive 0 enduse 14 # --
Re: Multi-tape question
On Tue, 26 Feb 2002, Don Potter wrote: What is the length of the tapetypeusing hardware compression you will need to adjust the length to what you expect your actually compression ratio to be..if you expect each tape to be about 70gb then your length would be 7 mbytes. Ah... so perhaps this is causing the issue? define tapetype DLTtapeIV { comment DLTtape IV - 40 GB length 33706 mbytes filemark 43 kbytes speed 1820 kps } That was what the tapetype run gave us. So I guess you're saying that using the cbn device rather than the bn or n device is creating a conflict? I'm just wondering why amanda would abandon partitions... it did seem to properly determine that it had ~70 Gb to work with... - Eric
Re: Multi-tape question
On Tue, 26 Feb 2002, Don Potter wrote: What is the length of the tapetypeusing hardware compression you will need to adjust the length to what you expect your actually compression ratio to be..if you expect each tape to be about 70gb then your length would be 7 mbytes. Ah... so perhaps this is causing the issue? define tapetype DLTtapeIV { comment DLTtape IV - 40 GB length 33706 mbytes filemark 43 kbytes speed 1820 kps } That was what the tapetype run gave us. So I guess you're saying that using the cbn device rather than the bn or n device is creating a conflict? I'm just wondering why amanda would abandon partitions... it did seem to properly determine that it had ~70 Gb to work with... - Eric
Re: Multi-tape question
On Tue, 26 Feb 2002, Don Potter wrote: When taper started writing the dumps to tape and it was determined that the filesystem exceeded the expected tape size (regardless of hardware compression) the dumps would fail since the dumps aren't capable of spanning tapes Waitaminit... this I didn't know. Amanda can't run dumps across tapes? Why is the runtapes entry even used at all? So if I use HW compression and set the length at 65 Gb, and I want to do a level 0 dump (say it's the first dump, for example) ... host/filesystem # 40 Gb FS host/filesystem2# 20 Gb FS host/filesystem3# 20 Gb FS ... Amanda will continue to sense a problem and drop one of the filesystems from the lineup? I thought that, using runtapes 2, amanda would put two of the filesystems on one tape, load the next tape, and stick the leftover system on it... ? - Eric
Re: Multi-tape question
On Tue, 26 Feb 2002, Don Potter wrote: It will put as many filesystems on a tape as possibel (prvoding tape length is calc'd correctly) and then write the remainder on the next tape (if you have specified the dumps to use multiple tapes) Understood. That's what I had believed going in, and fortunately, it's true. However, as you pointed out, amanda decided not to do that. Sorry for the confusion..I'll keep my mouth shut from now on Not at all! You told me what was wrong! I'm trying another run now with the original length and no compression so I can see if it works in UNDER three hours... :) - Eric
Resolved - Multi-tape question
Thank you to those that wrote back yesterday, especially Don. I used the non-compressed device (in this case, /dev/rmt/1n), and amanda did indeed write filesystems until it hit EOT and rewrote that specific interrupted filesystem to the next tape. Does anyone here have a changer working with more than one drive? The 30-slot StorEdge changer I'm working with has two drives installed, and I'm wondering how I can best utilize both of them. - Eric
Re: difference in tape devices question
On Tue, 5 Mar 2002, Davidson, Brian wrote: Can anyone tell me the difference between the following no-rewind tape devices on Linux? I can't seem to find any info anywhere /dev/nst0 /dev/nst0a /dev/nst0l /dev/nst0m I think the -l and -m notations instruct the system to send information in low and medium compressed formats, respectively I don't know what the -a means - Eric
Inconsistent taping
I just wanted to find out if any of you have seen this. I found out that I generally could get ~70 GB on a DLT-IV tape if I used hardware compression, so I set amanda's tape length to 60 GB and started backing up big partitions. I was able to backup 55 and 52 GB partitions, but there's one 50 GB partition is particular that won't fit on a tape, apparently. I've attempted it several nights in a row, but with no success, even when amanda starts the tape with it. Is what hardware compression can do with files variable? In this case, with the failure, it's a lot of oracle data, some of it is gzipped. - - - -- Eric Trager
Barcode reader
Hi all, I'm happily using chg-scsi with our StorEdge and backups are going well. Thanks to you all for helping me sort out a couple of things. I've tried to find an answer to this in the archives but to no avail. The StorEdge we use has a barcode reader, and the tapes are physically labelled. If I use mtx to check status, I get a nice list of tapes and barcode entries. I have an entry for labelfile in my changer config file, but when I run the chg-scsi -info command, I get 10 20 1 0 The fourth entry indicates I have no reader, but I do. How can I tell amanda and chg-scsi that there's a reader to use? - - - -- Eric Trager
Re: Barcode reader
On Mon, 11 Mar 2002, Thomas Hepper wrote: Hi, On Fri, Mar 08, 2002 at 11:38:32AM -0800, Stephen Carville wrote: [..] In changer.conf set havereader=1 Hmm i think it must be havebarcode 1 I'll try that. I actually already had havereader=1 in the conf but it didn't seem to make a difference. Which version of chg-scsi do you have. This option works not with 2.4.2 I have whatever version comes with 2.4.3b2 of Amanda. - - - -- Eric Trager
Re: Barcode reader
On Mon, 11 Mar 2002, Thomas Hepper wrote: Hi, On Fri, Mar 08, 2002 at 11:38:32AM -0800, Stephen Carville wrote: [..] In changer.conf set havereader=1 Hmm i think it must be havebarcode 1 Thomas, et al: I added this to the chg-scsi-solaris.conf file I use and the chg-scsi debug files now report that a barcode reader is available, but the labelfile for the configuration remains empty, even after running an amtape Config show. Any ideas on how to close the final gap on getting the mapping to work? Below is a debug file from a recent attempt to perform an amtape Config label LABEL. TIA, Eric chg-scsi: debug 1 pid 22862 ruid 0 euid 0 start time Mon Mar 11 16:21:29 2002 chg-scsi: $Id: chg-scsi.c,v 1.6.2.22.2.7.2.2 2001/12/30 17:26:22 martinea Exp $ ARG [0] : ./chg-scsi ARG [1] : -label ARG [2] : CMAG-12 Number of configurations: 1 Tapes need eject: Yes Inv. auto update: No barcode reader : Yes Tapes need sleep: 20 seconds Cleancycles : 1000 Changerdevice : /dev/changer Labelfile : /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Simple/labelfile Tapeconfig Nr: 0 Drivenumber : 0 Startslot : 0 Endslot : 19 Cleanslot : 30 Devicename: /dev/rmt/1bn changerident : none SCSITapedev : none tapeident : none statfile : /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Simple/tape1-status Slotfile : /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Simple/tape1-slot Cleanfile : /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Simple/tape1-clean Usagecount: /usr/cbel/opt/amanda-2.4.3b2/etc/amanda/Simple/totaltime DecodeModeSense : Element Address Assignment Page Medium Transport Element Address 1 Number of Medium Transport Elements 1 First Storage Element Address 256 Number of Storage Elements 30 First Import/Export Element Address 64 Number of ImportExport Elements1 First Data Transfer Element Address 128 Number of Data Transfer Elements 2 DecodeModeSense : MT can store data cartridges 1 DecodeModeSense : ST can store data cartridges 1 DecodeModeSense : IE can store data cartridges 1 DecodeModeSense : DT can store data cartridges 1 DecodeModeSense : MT to MT 0 DecodeModeSense : MT to ST 1 DecodeModeSense : MT to IE 1 DecodeModeSense : MT to DT 1 DecodeModeSense : ST to MT 1 DecodeModeSense : ST to MT 1 DecodeModeSense : ST to DT 1 DecodeModeSense : IE to MT 1 DecodeModeSense : IE to ST 1 DecodeModeSense : IE to ST 1 DecodeModeSense : IE to ST 1 DecodeModeSense : DT to MT 1 DecodeModeSense : DT to ST 1 DecodeModeSense : DT to IE 1 DecodeModeSense : DT to DT 1 ChgExit in MapBarCode, reason MapBarCode, DB Version does not match
Re: Barcode reader
On Sun, 17 Mar 2002, Thomas Hepper wrote: Hi, On Fri, Mar 15, 2002 at 10:48:24AM -0500, Eric Trager wrote: I added this to the chg-scsi-solaris.conf file I use and the chg-scsi [...] ChgExit in MapBarCode, reason MapBarCode, DB Version does not match You should drop the old labelfile and rerun an amtape conf show I did and unfortunately, while the show ran, a labelfile was not created. I also followed John's suggestion and ran an update but with no result. However, I did reset the changer and run ./chg-scsi -label TEMP-00 which was the label of the tape in the drive at the time. I got back 0 0 0 and the labelfile was created with an entry :2:TEMP-00 So I ran a amtape Temp slot next, which loaded the next tape, TEMP-01, and then ran the label command ./chg-scsi -label TEMP-01 and the labelfile changed, but the entry wrote over the first one: :2:TEMP-01 was all that was in the labelfile. I'm not sure what the 2 is, or what the fields are supposed to be. Just to show what the barcodes are like, here is a line from mtx: Storage Element 5:Full :VolumeTag=A4 What do you all think? Should I just throw in the towel? I feel like I'm getting there, but the road is all over the place. - Eric PS: Here are some lines from the debug file on the amtape Temp update command I ran earlier: changer: got exit: 0 str: 1 20 1 0 changer: got exit: 0 str: 1 /dev/rmt/1bn changer_label: TEMP-01 for slot 1 changer: got exit: 0 str: 1 20 1 0 changer_query: changer return was 20 1 0 changer_query: searchable = 0 changer: got exit: 0 str: 2 /dev/rmt/1bn changer_label: TEMP-02 for slot 2 [etc.]
Re: Tar 1.13.25?
On Mon, 18 Mar 2002, Morse, Richard E. wrote: Hi! I've seen on FreeBSD a listing for gtar 1.13.25. However, I can't seem to find the sub-revision number in the sources downloaded from ftp.gnu.org , and the most recent rpm from redhat is for 1.13.19-1. Are the sources on ftp.gnu.org for 1.13.25? Or should I just upgrade the lone gnu/linux machine 1.13.19? I had trouble finding 1.13.19 when I needed it. I ended up getting it from ftp://alpha.gnu.org/gnu/tar/. - Eric
Re: Two tape drives
On Mon, 1 Apr 2002, C. Chan wrote: Is there currently a way to back up to two or more tape drives simultaneously using the same Amanda config? Apparently not. I have a 30-slot jukebox attached to two drives and use two different configs to run the robot and perform backups (sometimes simultaneously), but amanda can't run one backup across two tapes in two drives in parallel. We have another system with a four-drive jukebox and Legato Networker and Networker seems to handle the parallel stuff pretty well... for a price. - - - -- Eric Trager CMAG, DCB, CIT, NIH 12A/2020, 301-402-8738 [EMAIL PROTECTED]
Re: disklist and hw compression
On Tue, 2 Apr 2002, Matthew Boeckman wrote: I'm running an ultra5, and trying to level0 a 60+gig filesystem. Using software compression took over 22 hours (i killed it at that point) so I decided to use hardware, specifying /dev/rmt/0cn as the tape device in amanda.conf. When I tried to run amdump, it failed saying that dumps too large for disk, which is true if you don't count the hardware compression I was trying to use. The tapes are AIT2 50GB tapes. Is there a way to force amanda to do this? I assume perhaps amadmin csd force diskname ? I wanted to know for sure before kicking off another 20+hour backup cycle. Or is there something that I have to configure in amanda to let it know i'm using hardware compression and to ignore tape warnings? Make sure that you configure the tape length in amanda.conf to specify what you think it is that you'll be able to get on a tape counting hardware compression. When I used (note past tense) hardware compression with our systems, I was using 40 GB DLT tapes. With regular, unzipped data I was getting close to 70 GB on a tape, so I set the tape length to 65000 Mb. It worked fine with several partitions and combinations, anywhere from 45 to 65 GB in size. Unfortunately, one particular partition, only 45-50 GB in size, was loaded with small directories containing nothing but gzips. I was not able to get that on a tape until I used GNUTAR to break it up. I stopped using hardware compression because the general attitude among the experts here was that amanda is better off doing her own packing. I haven't really noticed a significant difference, but, as always, YMMV, etc. - - - -- Eric Trager
Re: Documentation
On 19 Apr 2002, Marcos Teodoro Dias Junior wrote: Hello, Where I can find some documentation at how to use amanda with tape changers ? In the source tree, look in docs/TAPE.CHANGERS . - Eric
Re: ADIC / Dell Powervault 120T changer problems under Linux (Pt 2)
On Sun, 24 Mar 2002, Derrick Miller wrote: So, my questions: 1) Does anyone have this changer working with Amanda? I would give just about anything to get my hands on copies of working config and chg-zd-mtx files. For what it may be worth, Derrick, I have a Sun StorEdge jukebox attached to an Enterprise class Solaris system. While mtx itself worked, I never got the chg-zd-mtx script to work. I had to do things like hard-code variables just to even get it to run but I never got it to perform any significant functions. I inhereted the installation and found out that chg-scsi was never tried, but lo and behold, I spent a little time tweaking it, it worked well, and have used it ever since. It really is a great little program. The only functionality that does not work for me right now is the barcode stuff but Thomas Hepper has been working with me to make that work. BTW, I use 2.4.3b2. The StorEdge is a 30-slot jukebox with two DLT IV tape drives. Any pointers whatsoever would be greatly appreciated, as I've been beating my head against this for over a month and have made very little progress. Good luck. If I hadn't discovered chg-scsi would do the job, I would have tried writing my own perl script. Lots of people use chg-zd-mtx, though, which always made me feel worse! :^P - - - -- Eric Trager CMAG, DCB, CIT, NIH [EMAIL PROTECTED]