Re: How to fix annoying break in tape sequence?
On Monday, 01.12.2003 at 12:51 -0500, Brian Cuttler wrote: > > Remember to set a calendar reminder to reactivate it. > > We always run amcheck -m from cron - if the tapes are out of sequence > we notice (also someone reads the amdump output which lists both > current and next tapes). Already doing both of the above :-) Dave. -- Dave Ewart [EMAIL PROTECTED] Computing Manager, Epidemiology Unit, Oxford Cancer Research UK PGP: CC70 1883 BD92 E665 B840 118B 6E94 2CFD 694D E370
Re: How to fix annoying break in tape sequence?
On Monday, 01.12.2003 at 20:11 +0100, Gerhard den Hollander wrote: > > We have a four-week, five days per week cycle: > > Good :) > > > These twenty tapes are named OurName-A-Mon, OurName-A-Tue, ..., > > OurName-A-Fri, OurName-B-Mon, ..., OurName-D-Fri (in other words, > > the letters A, B, C, D refer to weeks in the four-week cycle): > > Bad idea. > > Why not simpl;y name the tapes YourName001 -> YourName020 Yes, I know you're right. To be honest, that is something I'm considering changing. I have a method sorted out to do gradually rename all the tapes. > And then every day you have a cronjob (say at 15:00 and at 16:00) that > does an amcheck -m YourConfig We do that anyway, just in case we've forgotten to load the tape. "Knowing" what the next tape is, is no more/less easy regardless of whether you use YourName001 -> YourName020, or OurName-A-Mon etc. > When you come to Xmas, you will either have to go to work to replace > the tapes, or have to live with a shifted week. Actually, this one's easy, because we have a clean two-week shutdown at Christmas. :-) Dave. -- Dave Ewart [EMAIL PROTECTED] Computing Manager, Epidemiology Unit, Oxford Cancer Research UK PGP: CC70 1883 BD92 E665 B840 118B 6E94 2CFD 694D E370
Tapetype Overland LoaderExpress with HP Ultrium 1 (LXL1U11)
Hello, here is the Tapetype-Definition for the Overland LoaderExpress with HP Ultrium 1 device 11-Slot-Loader. It's the output from the tapetype-Run (after 33 hours :-(). define tapetype OVERLAND-LXL1U11 { comment "OVERLAND AutoLoader HP Ultrium 1" length 104251 mbytes filemark 537 kbytes speed 1736 kps } The Loader runs fine with the mtx-Tools and Configuration. Greetings Bernd
Re: Tapetype Overland LoaderExpress with HP Ultrium 1 (LXL1U11)
Postaremczak Bernd wrote: Hello, here is the Tapetype-Definition for the Overland LoaderExpress with HP Ultrium 1 device 11-Slot-Loader. It's the output from the tapetype-Run (after 33 hours :-(). It would have been much faster (about 6-7 hours) if you used the amtapetype from 2.4.4 and specified a reasonable estimate, just like in it's manpage. Something like: amtapetype -e 100g define tapetype OVERLAND-LXL1U11 { comment "OVERLAND AutoLoader HP Ultrium 1" length 104251 mbytes filemark 537 kbytes speed 1736 kps } Capacity is correct, but speed is a little slow (due to not giving a realistic estimate). -- Paul @ Home
RE: Tapetype Overland LoaderExpress with HP Ultrium 1 (LXL1U11)
> Postaremczak Bernd wrote: > > Hello, > > > > here is the Tapetype-Definition for the Overland > LoaderExpress with HP > > Ultrium 1 device 11-Slot-Loader. It's the output from the > tapetype-Run > > (after 33 hours :-(). > > It would have been much faster (about 6-7 hours) if you used the > amtapetype from 2.4.4 and specified a reasonable estimate, just like > in it's manpage. Something like: amtapetype -e 100g I'll try it next time. > > define tapetype OVERLAND-LXL1U11 { > > comment "OVERLAND AutoLoader HP Ultrium 1" > > length 104251 mbytes > > filemark 537 kbytes > > speed 1736 kps > > } > > Capacity is correct, but speed is a little slow (due to not giving > a realistic estimate). Maybe it is the SCSI-Controller. He works only with 20 MBytes/sec. Have you got better definitions for it? Bernd
SendSize CoreDmp on Aix 4.3.3
Hello, I am using amanda 2.4.4p1 on a AIX 4.3.3 IBM RS6000. I have compiled it with both, the ibm coompiler en gcc 2.95.3. The both gave me an unusable sendsize command. This command end in a coredmp. For some obscure raison the executable crash on the call to amroflock, dont know why (seem to be a link error ..) Well after 1 day, i found a solution. I recompiled all amanda with --disable-shared option, and after that all seems to work normaly Hope it can help others ... Regards Fabrice
Re: SendSize CoreDmp on Aix 4.3.3
Didierjean Fabrice wrote: Hello, I am using amanda 2.4.4p1 on a AIX 4.3.3 IBM RS6000. I have compiled it with both, the ibm coompiler en gcc 2.95.3. The both gave me an unusable sendsize command. This command end in a coredmp. For some obscure raison the executable crash on the call to amroflock, dont know why (seem to be a link error ..) Well after 1 day, i found a solution. I recompiled all amanda with --disable-shared option, and after that all seems to work normaly Hope it can help others ... It sure does ! I was hoping for months to be able to use amanda on my old AIX server but I had the same problem as you. I scrached my old amanda compilation, retrieved a fresh one, compiled it with the --disable-shared option and after some tests, it does works very nice. Thank you very much for that information. -- Nicolas Ecarnot
RE: Help - recovering without amanda
> I always restore to a temp space, just in case. I should have done this...I never do direct restores...ugh > > > # mt rewind > > # mt fsf 1 > > # dd if=/dev/rmt/0hn bs=1 skip=1 | /usr/local/bin/tar -xf - > ^ > Bad. It should be bs=32k, as above (and as in the header you > got running > the first command). Also, an even safer way is to not do the pipe: > > dd if=/dev/rmt/0hn bs=32k skip=1 of=output.file > > Then you can do 'tar t' on the output file to get a table of > contents and *really* make sure it's what you want. The bs=1 was a typo...I can't cut and paste because this box is on a test LAN that's not even getting out the door...my bad. But skipping the pipe is a good suggestion. > > Here's a brief summary. Amanda stores several files on a > tape. The first > is the tape header. That's what you skip over with 'mt fsf > 1'. The next > file is the first dump image. The next is the second image, > etc. Each > dump image has a 32k amanda header, and then the image. > > The dd command with with 'bs=32k skip=1' reads the whole > file, skipping > over the first 32k -- the amanda header. It stops when it > hits EOF of > that dump image. If you run the exact same dd command again, > you'll grab > the next backup image. > Thanks...I'll keep plugging away...
Testing Tape Drive
I'd like to run an new tape drive through its paces to make sure it is working full before I start trying to learn AMANDA on it. What are some commands that i can use to write, read and generally get familiar with it and be sure it is working as desired? It is a Compaq SDT-9000 drive on Gentoo Linux (which requires devfs) and so it probably is in a non-standard area. I've been reading around and have taken a lot in, but seem to be having trouble organizing the information in my head. If someone would be willing to just punch out some commands to help my brain stitch together what it's picked up I'd greatly appreciate it. I know mt and tar should be used. Thanks, JSR/ -- System Administrator Washington Bible College/Capital Bible Seminary http://www.bible.edu
Re: SendSize CoreDmp on Aix 4.3.3
On Tue, Dec 02, 2003 at 11:39:39AM +0100, Didierjean Fabrice wrote: > Hello, > > I am using amanda 2.4.4p1 on a AIX 4.3.3 IBM RS6000. I have compiled it > with both, the ibm coompiler en gcc 2.95.3. The both gave me an unusable > sendsize command. This command end in a coredmp. For some obscure raison > the executable crash on the call to amroflock, dont know why (seem to be > a link error ..) > > Well after 1 day, i found a solution. I recompiled all amanda with > --disable-shared option, and after that all seems to work normaly That sounds like a shared library is seen at compile-time but not at run-time. The "link error" you note (without the --disable-shared option) is probably the run-time "dynamic linker" not finding the library that the compiler did find. Shared libraries are preferable due to better use of memory and disk space. One copy of the shared library for all programs that use it as opposed to a separate, private copy, in each program. Think about how many C programs use "printf" and what a waste it would be if each has a separate, private copy. I scanned some AIX documentation online - meaning I could have, and most likely did, miss a lot :(( I'll relate some of what I found to Solaris which I know better. Maybe someone who knows AIX better can correct any omissions or errors. Most compilers and linkers use a "library search path" to locate the libraries they need. There is generally a default set of directories to search and this can be extended in several ways. One way, generally considered the preferable way, is with compile/ linker options. On Solaris some relevant options include: -l # a library that will be needed -L# a directory for the compiler to search in -R# a directory for the runtime linker to search Note, on Solaris the search path is extended separatly for the compiler and the runtime linker. My scan of AIX documentation suggests it only has the "-L" and that it may serve both purposes. A second, less prefered way, to extend the library search path is with an environment variable. On Solaris the appropriate variable is "LD_LIBRARY_PATH". My reading of AIX stuff suggests the compareable variable there is "LIBPATH". The setting of this variable could be a source of your troubles when using shared libs. It may be set one way during compile, but when amanda runs it from cron it may be different. That is one reason for prefering to set the compile and runtime paths in the executable with the -L (and maybe -R) options. Solaris also has one command that I did not find a comparable one on AIX. That is the "ldd" command. It looks at an executable file and lists the shared libraries it needs and where it finds them (or doesn't find them). Here is an example: $ ldd firebird-bin libdl.so.1 =>/usr/lib/libdl.so.1 libglib-1.2.so.0 => /usr/sfw/lib/libglib-1.2.so.0 libc.so.1 => /usr/lib/libc.so.1 libnsl.so.1 => /usr/lib/libnsl.so.1 libmozjs.so => (file not found) libX11.so.4 => /usr/lib/libX11.so.4 libsocket.so.1 =>/usr/lib/libsocket.so.1 libm.so.1 => /usr/lib/libm.so.1 libCstd.so.1 => /usr/lib/libCstd.so.1 libxpcom.so => (file not found) libpthread.so.1 => /usr/lib/libpthread.so.1 libgtk-1.2.so.0 => /usr/sfw/lib/libgtk-1.2.so.0 This output says most of the libs will be found at run-time. But two libs (libmozjs.so and libxpcom.so) were found properly during compile, but at run-time they will not be found. This executable may run ok until something in the "not found" libraries is needed. If I've compiled this executable I go back and set the appropriate options so the run-time linker finds the libraries. Otherwise, as above where I did not compile the program, I run it from a wrapper shell that first sets LD_LIBRARY_PATH (your LIBPATH) appropriately and then calls the actual executable. As I noted, I did not find a program compareable to ldd in my scan of AIX docs. There probably is one. It would be a good diagnostic tool to put into your toolkit. jl -- Jon H. LaBadie [EMAIL PROTECTED] JG Computing 4455 Province Line Road(609) 252-0159 Princeton, NJ 08540-4322 (609) 683-7220 (fax)
Re: Testing Tape Drive
On Tue, Dec 02, 2003 at 09:34:23AM -0500, Josiah Ritchie wrote: > I'd like to run an new tape drive through its paces to make sure it is working > full before I start trying to learn AMANDA on it. What are some commands that i > can use to write, read and generally get familiar with it and be sure it is > working as desired? It is a Compaq SDT-9000 drive on Gentoo Linux (which > requires devfs) and so it probably is in a non-standard area. > > I've been reading around and have taken a lot in, but seem to be having trouble > organizing the information in my head. If someone would be willing to just punch > out some commands to help my brain stitch together what it's picked up I'd > greatly appreciate it. I know mt and tar should be used. Unless that is a changer you pretty much have it. Write some write a tar archive to the tape and be sure you can recover it. Then write multiple archives to the tape using the "no rewind" version of the tape device. Use mt to see if you can position at a specific archive and recover just the one. You might want to use dd with a blocksize of 32K to be sure the default amanda block size works ok. Similarly you might want to explore the mt commands to set compression on/off, block size, check status, set defaults, ... If it is a changer, also explore the mtx command. -- Jon H. LaBadie [EMAIL PROTECTED] JG Computing 4455 Province Line Road(609) 252-0159 Princeton, NJ 08540-4322 (609) 683-7220 (fax)
su hangs vs 2.6.x kernels, problem found & fixed
Greetings everybody; As the transition to the 2.6 kernel family is eminent, there is one minor gotcha that I can tell you howto fix. Doing an 'su amanda' in order to do the install hangs the su process occasionally. If you search thru the process table tree, it can be seen that su has started another copy of bash, and that bash has a copy of stty linked to it. Everything above stty is sleeping, waiting on stty to init I guess. However stty is shown as stopped, so the shell is well and truely hung. Note that this doesn't seem to effect the 'su amanda -c "command"' version of the su usage, only that which actually returns you a shell with amanda rights seems to be effected. It turns out there is a missing call in the readymade versions of bash being shipped with most distro's. Its called PRGP_PIPE in the .config. The fedora release of redhat contains a bash rpm that fixes this. It installed with no hiccups here on this RH8.0 system, and solves that particular problem. The test to see if the call is missing in your installed version of bash is: #>strings /bin/bash|grep pgrp_pipe If this returns a null, then its missing and you'll need to find a bash that has it, or rebuild yours (shudder) after editing the file config.h to add this line: - #define PGRP_PIPE 1 - I hope this helps the smooth transition to 2.6. -- Cheers, Gene AMD [EMAIL PROTECTED] 320M [EMAIL PROTECTED] 512M 99.27% setiathome rank, not too shabby for a WV hillbilly Yahoo.com attornies please note, additions to this message by Gene Heskett are: Copyright 2003 by Maurice Eugene Heskett, all rights reserved.
Re: Testing Tape Drive
Jon LaBadie scripted :: >Unless that is a changer you pretty much have it. Write some >write a tar archive to the tape and be sure you can recover it. >Then write multiple archives to the tape using the "no rewind" >version of the tape device. Use mt to see if you can position >at a specific archive and recover just the one. > >You might want to use dd with a blocksize of 32K to be sure >the default amanda block size works ok. Similarly you might >want to explore the mt commands to set compression on/off, >block size, check status, set defaults, ... > >If it is a changer, also explore the mtx command. It isn't a changer so I can skip the mtx curve. :-) I'll probably need to get into that later though cause I'll eventually need one. I can't play around on this server too much for fear of messing things up. Can I pass some commands by you guys for sanity checking? mt -f /dev/st0 status # works fine (If I know what I'm reading) mt -f /dev/st0 retension # works mt -f /dev/st0 erase # works I need to remember a & at the end next time, ssh won't background it for some odd reason. (^z bg) tar -cf /dev/st0 /etc #? records /etc to tape in /dev/st0 JSR/
Re: SendSize CoreDmp on Aix 4.3.3
Jon LaBadie wrote: That sounds like a shared library is seen at compile-time but not at run-time. The "link error" you note (without the --disable-shared option) is probably the run-time "dynamic linker" not finding the library that the compiler did find. There was no error at run time .. only a core dump. On my systeme there was the ldd commande and when i use it on the executable it says it found all the libs. I have lauch gdb on the exe and in fact on the line 92 of the file amandates.c it calls the function amroflock : rc = amroflock(fileno(amdf), "amandates"); This function amroflock is in the file amflock.c between the line 319 and 314, or the call (trace with gdb) falls at the line 398, which is the last line of the function amfunlock ... I have try the operation whit gcc and with ibm compilator .. the two give the same error. Here is the result of an execution : bash-2.04# pwd /opt-net/amanda/AIX/libexec bash-2.04# ldd sendsize /opt-net/amanda/AIX/lib/libamclient.a(libamclient-2.4.4p1.so) /usr/lib/libintl.a(shr.o) /opt-net/amanda/AIX/lib/libamanda.a(libamanda-2.4.4p1.so) /usr/lib/libc.a(pse.o) /usr/lib/libtli.a(shr.o) /usr/lib/libpthreads.a(shr.o) /usr/lib/libpthreads_compat.a(shr.o) /usr/lib/libnsl.a(shr.o) /usr/lib/libpthreads.a(shr_xpg5.o) /usr/lib/libpthreads.a(shr_comm.o) sendsize /usr/lib/libcrypt.a(shr.o) /usr/lib/libc.a(shr.o) bash-2.04# ./sendsize Segmentation fault (core dumped) bash-2.04# gdb ./sendsize GNU gdb 5.0 Copyright 2000 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "powerpc-ibm-aix4.3.2.0"... (gdb) run Starting program: /opt-net/amanda/AIX/libexec/./sendsize Program received signal SIGSEGV, Segmentation fault. 0xd0dee728 in amroflock () at /projets/adminsys/src/amanda-2.4.4p1/common-src/amflock.c:398 398 } (gdb) bt #0 0xd0dee728 in amroflock () at /projets/adminsys/src/amanda-2.4.4p1/common-src/amflock.c:398 #1 0xd0dece4c in start_amandates (open_readwrite=0) at /projets/adminsys/src/amanda-2.4.4p1/client-src/amandates.c:92 #2 0x10003940 in main (argc=1, argv=0x2ff22ba8) at /projets/adminsys/src/amanda-2.4.4p1/client-src/sendsize.c:159 #3 0x11dc in __start () (gdb) quit The program is running. Exit anyway? (y or n) y bash-2.04# [snip] regards Fabrice
Backup taking more than 2 days to complete, help!
Hi, since one week ago, my daily backups are taking more than 2 days to complete! I dont know why yet, but just one directory in a client machine (host2:/var/vpopmail) is taking a lot of time, as seen in the output of the amstatus below. Here is the amanda proccess right now running in the client machine: amanda 32319 0.0 0.1 1552 436 ?? SMon07AM 0:12.50 /usr/local/amanda/libexec/sendbackup amanda 32320 0.0 0.1 1552 352 ?? SMon07AM 4:47.62 /usr/local/amanda/libexec/sendbackup amanda 32322 0.0 0.0 6280 ?? IW - 0:00.00 sh -c /usr/local/bin/tar -tf - 2>/dev/null | sed -e 's/^\\.//' amanda 32323 0.0 0.1 1068 372 ?? SMon07AM 2:47.97 /usr/local/bin/tar -tf - amanda 32324 0.0 0.1 972 368 ?? SMon07AM 0:10.05 sed -e s/^\\.// Im not using compression neither in the server or the client. host2:/var/vpopmail contains 2.4 GB of data, with 168577 files. The strange is that it begins to happen with no reason My guess is that gnu tar is eating all the time... AMSTATUS OUTPUT: host1:/etc01184k finished (0:09:56) host1:/root 0 32k finished (0:09:10) host1:/usr/adm0 29440k finished (0:10:48) host1:/usr/home 0 32k finished (0:09:28) host1:/usr/local 0 169920k finished (0:14:27) host1:/var/qmail 01184k finished (0:09:44) host2:/etc 01120k finished (0:11:01) host2:/root 0 91936k finished (1:06:02) host2:/usr/home 0 32k finished (0:10:00) host2:/usr/local0 701824k finished (8:50:05) host2:/usr/share0 94560k finished (2:00:00) host2:/var/qmail/control 0 32k finished (0:09:25) host2:/var/qmail/supervise 0 32k finished (0:09:14) host2:/var/qmail/users 0 32k finished (0:09:40) host2:/var/vpopmail 0 4473120k finished (2+5:47:06) SUMMARY part real estimated size size partition : 15 estimated : 155961992k failed : 0 0k ( 0.00%) wait for dumping: 0 0k ( 0.00%) dumping to tape : 0 0k ( 0.00%) dumping : 00k0k ( 0.00%) ( 0.00%) dumped : 15 5564480k 5961992k ( 93.33%) ( 93.33%) wait for writing: 00k0k ( 0.00%) ( 0.00%) writing to tape : 00k0k ( 0.00%) ( 0.00%) failed to tape : 00k0k ( 0.00%) ( 0.00%) taped : 15 5564480k 5961992k ( 93.33%) ( 93.33%) 4 dumpers idle : not-idle taper idle network free kps: 1900 holding space : 11918382k (100.00%) dumper0 busy : 2+4:36:05 ( 98.08%) dumper1 busy : 0:00:00 ( 0.00%) taper busy : 1:16:16 ( 2.37%) 0 dumpers busy : 1:01:52 ( 1.92%)not-idle: 1:00:56 ( 98.50%) start-wait: 0:00:55 ( 1.50%) 1 dumper busy : 2+4:36:06 ( 98.08%)not-idle: 1+20:05:42 ( 83.83%) client-constrained: 8:27:35 ( 16.08%) no-bandwidth: 0:02:00 ( 0.06%) start-wait: 0:00:47 ( 0.03%) 2 dumpers busy : 0:00:00 ( 0.00%)
Re: Backup taking more than 2 days to complete, help!
--On Tuesday, December 02, 2003 15:46:57 -0200 Fernando Costa de Almeida <[EMAIL PROTECTED]> wrote: > Hi, > > since one week ago, my daily backups are taking more than 2 days to > complete! I dont know why yet, but just one directory in a client > machine (host2:/var/vpopmail) is taking a lot of time, as seen in the > output of the amstatus below. Here is the amanda proccess right now > running in the client machine: > > amanda 32319 0.0 0.1 1552 436 ?? SMon07AM 0:12.50 > /usr/local/amanda/libexec/sendbackup > amanda 32320 0.0 0.1 1552 352 ?? SMon07AM 4:47.62 > /usr/local/amanda/libexec/sendbackup > amanda 32322 0.0 0.0 6280 ?? IW - 0:00.00 sh -c > /usr/local/bin/tar -tf - 2>/dev/null | sed -e 's/^\\.//' > amanda 32323 0.0 0.1 1068 372 ?? SMon07AM 2:47.97 > /usr/local/bin/tar -tf - > amanda 32324 0.0 0.1 972 368 ?? SMon07AM 0:10.05 sed -e > s/^\\.// > > Im not using compression neither in the server or the client. > host2:/var/vpopmail contains 2.4 GB of data, with 168577 files. > > The strange is that it begins to happen with no reason > My guess is that gnu tar is eating all the time... > > AMSTATUS OUTPUT: > > host1:/etc01184k finished (0:09:56) > host1:/root 0 32k finished (0:09:10) > host1:/usr/adm0 29440k finished (0:10:48) > host1:/usr/home 0 32k finished (0:09:28) > host1:/usr/local 0 169920k finished (0:14:27) > host1:/var/qmail 01184k finished (0:09:44) > host2:/etc 01120k finished (0:11:01) > host2:/root 0 91936k finished (1:06:02) > host2:/usr/home 0 32k finished (0:10:00) > host2:/usr/local0 701824k finished (8:50:05) > host2:/usr/share0 94560k finished (2:00:00) > host2:/var/qmail/control 0 32k finished (0:09:25) > host2:/var/qmail/supervise 0 32k finished (0:09:14) > host2:/var/qmail/users 0 32k finished (0:09:40) > host2:/var/vpopmail 0 4473120k finished (2+5:47:06) If host2 is not your Amanda server, my bet would be a network problem, probably a duplex mismatch. Can you ftp a large file from host2 to your Amanda server and get reasonable a transfer rate? Frank > > SUMMARY part real estimated > size size > partition : 15 > estimated : 155961992k > failed : 0 0k ( 0.00%) > wait for dumping: 0 0k ( 0.00%) > dumping to tape : 0 0k ( 0.00%) > dumping : 00k0k ( 0.00%) ( 0.00%) > dumped : 15 5564480k 5961992k ( 93.33%) ( 93.33%) > wait for writing: 00k0k ( 0.00%) ( 0.00%) > writing to tape : 00k0k ( 0.00%) ( 0.00%) > failed to tape : 00k0k ( 0.00%) ( 0.00%) > taped : 15 5564480k 5961992k ( 93.33%) ( 93.33%) > 4 dumpers idle : not-idle > taper idle > network free kps: 1900 > holding space : 11918382k (100.00%) > dumper0 busy : 2+4:36:05 ( 98.08%) > dumper1 busy : 0:00:00 ( 0.00%) >taper busy : 1:16:16 ( 2.37%) > 0 dumpers busy : 1:01:52 ( 1.92%)not-idle: 1:00:56 ( > 98.50%) >start-wait: 0:00:55 ( > 1.50%) > 1 dumper busy : 2+4:36:06 ( 98.08%)not-idle: 1+20:05:42 > ( 83.83%) > client-constrained: 8:27:35 ( > 16.08%) > no-bandwidth: 0:02:00 ( > 0.06%) > start-wait: 0:00:47 ( > 0.03%) > 2 dumpers busy : 0:00:00 ( 0.00%) > > -- Frank Smith [EMAIL PROTECTED] Systems Administrator Voice: 512-374-4673 Hoover's Online Fax: 512-374-4501
Re: Testing Tape Drive
--On Tuesday, December 02, 2003 12:35:06 -0500 Josiah Ritchie <[EMAIL PROTECTED]> wrote: > Jon LaBadie scripted :: > >> Unless that is a changer you pretty much have it. Write some >> write a tar archive to the tape and be sure you can recover it. >> Then write multiple archives to the tape using the "no rewind" >> version of the tape device. Use mt to see if you can position >> at a specific archive and recover just the one. >> >> You might want to use dd with a blocksize of 32K to be sure >> the default amanda block size works ok. Similarly you might >> want to explore the mt commands to set compression on/off, >> block size, check status, set defaults, ... >> >> If it is a changer, also explore the mtx command. > > It isn't a changer so I can skip the mtx curve. :-) I'll probably need to get > into that later though cause I'll eventually need one. > > I can't play around on this server too much for fear of messing things up. Can I > pass some commands by you guys for sanity checking? > > mt -f /dev/st0 status # works fine (If I know what I'm reading) > mt -f /dev/st0 retension # works > mt -f /dev/st0 erase # works I need to remember a & at the end next time, ssh > won't background it for some odd reason. (^z bg) > > tar -cf /dev/st0 /etc #? records /etc to tape in /dev/st0 Also make sure you can then mt -f /dev/st0 rewind tar -xf /dev/st0 (first make sure you are using GNU tar and are in a temp directory so you don't clobber your /etc) As Jon pointed out, you need to make sure you can use the proper non-rewinding device: mt -f /dev/nst0 rewind tar -cv /dev/nst0 /some/dir tar -cf /dev/nst0 /another dir mt -f /dev/nst0 rewind tar -tf /dev/nst0 (should list /some/dir) mt -f /dev/nst0 fsf 1 (moves to beginning of next file) tar -tf /dev/nst0 (should list /another/dir mt -f /dev/st0 rewind If you're planning on using tar for backups: tar --version Make sure it is at least 1.13.19, I think 1.13.25 is the current version. Older versions will appear to back up fine but you will have problems restoring properly. > > JSR/ -- Frank Smith [EMAIL PROTECTED] Systems Administrator Voice: 512-374-4673 Hoover's Online Fax: 512-374-4501
Re: Backup taking more than 2 days to complete, help!
On Tue, Dec 02, 2003 at 12:22:26PM -0600, Frank Smith wrote: > --On Tuesday, December 02, 2003 15:46:57 -0200 Fernando Costa de Almeida <[EMAIL > PROTECTED]> wrote: > > > Hi, > > > > since one week ago, my daily backups are taking more than 2 days to > > complete! I dont know why yet, but just one directory in a client > > machine (host2:/var/vpopmail) is taking a lot of time, as seen in the > > output of the amstatus below. Here is the amanda proccess right now > > running in the client machine: It is not unique in the rate of backup. It is large and thus takes a lot of time. But you have 4 DLE's on that host with greater than 2 MB. Each has the same backup rate, about 28KB/sec. So it is not the directory, most likely as Frank suggests a network problem. > > > > Im not using compression neither in the server or the client. > > host2:/var/vpopmail contains 2.4 GB of data, with 168577 files. Maybe now, but it had 4.4GB when it was backed up. > > > > The strange is that it begins to happen with no reason > > My guess is that gnu tar is eating all the time... Doubtful. Well, maybe if it is a cygwin client on a PC. > > > > AMSTATUS OUTPUT: ... > > > > host2:/root 0 91936k finished (1:06:02) > > host2:/usr/local0 701824k finished (8:50:05) > > host2:/usr/share0 94560k finished (2:00:00) > > host2:/var/vpopmail 0 4473120k finished (2+5:47:06) These are the DLE's, all with the same dump rate. Likely network limited. > > If host2 is not your Amanda server, my bet would be a network > problem, probably a duplex mismatch. Can you ftp a large file > from host2 to your Amanda server and get reasonable a transfer > rate? > > Frank jl -- Jon H. LaBadie [EMAIL PROTECTED] JG Computing 4455 Province Line Road(609) 252-0159 Princeton, NJ 08540-4322 (609) 683-7220 (fax)
Is it a bad idea to share the /holdingdisk/ area among several concurrent dump sets?
Hi -- [I'm _finally_ getting this posted! I've been trying for days to submit it through Yahoo, but it failed every time. Thanks for telling me about the Yahoo interface being broken and reminding me about the mailing list, Jon L!! :-] I did a search through the FAQ-O-Matic and through the Yahoo egroup for 'share holding' and didn't find anything specific in response to this question, so I think it's safe to ask: Is it a bad idea to have a central /holdingdisk/ area in use by multiple concurrent dump sets? But before you answer that, maybe I should ask this: Is it a bad idea to run multiple concurrent dump sets in the first place? (While searching the egroup archive for an answer to this, I came across comments from some of the Amanda veterans that led me to believe that I should be using a _single_ backup set, instead of running multiple concurrent ones. Am I reading that correctly? If so, then the original question is moot...) I've attached one of my configuration files (amanda4.conf.txt) with all its include files, the chg-scsi.conf(.txt) file, and its disklist (disklist4.txt). (I had to rename all the files to end with ".txt" so that they show up as text files. *sigh*) This particular dump set configuration is different than the other three I run concurrently. For each of those, a full dump is about 20gb, but this one is for some much larger filesystems. I've done a one-time full backup, and now I'm trying to set it up to only do incrementals now, but even doing that, I think it's going to take 2-3 35gb tapes to hold each night's incremental dumps. But isn't working anyway. It gets to a certain point and then _everything_ stops. I don't see any incrementing for the various filehandles (checked via 'lsof') -- not taper's tape device filehandle nor any of the sendbackup->dumper pipes. (If I do need to consolidate the four dump sets into one, I wonder what it'll take to make this one play nicely with the other three...) Separate question: I started to look at 2.4.4p1, but I ran into an issue right off the bat. With 2.4.4, 'configure' reported "fcntl locking works... yes", but with 2.4.4p1, I get this: checking whether posix fcntl locking works... no checking whether flock locking works... no checking whether lockf locking works... no checking whether lnlock locking works... no configure: WARNING: *** No working file locking capability found! configure: WARNING: *** Be VERY VERY careful. (I get the same results whether I build with ANSI C or with GCC.) I don't see any mention in the NEWS or ChangeLog files of any changes in this area. I do have some degree of parallelism in my configuration, but I can't tell if there've been any situations up to now where file locking has been needed and used, so I'm reluctant to go any further with 2.4.4p1. Has anyone else run into this? Thanks! -- Mark Conty Cargill, Inc. # Configuration file for chg-scsi tape changer program. # number_configs 4 #emubarcode 0 autoinv 1 havebarcode 1 debuglevel 9:0 eject 0 sleep 1 changerdev /dev/picker labelfile /opt/amanda/etc/labelfile usagecount /opt/amanda/etc/totaltime # Drive-specific configurations: # config 0 drivenum0 dev /dev/rmt/c4t1d0NOCOMPn # aka /dev/rmt/1mn startuse0 # chg-scsi uses 0-based numbering enduse 9 statfile/opt/amanda/etc/daily1/tape-slot tapestatus /opt/amanda/etc/daily1/tape-status cleanfile /opt/amanda/etc/daily1/tape-clean config 1 drivenum1 dev /dev/rmt/c4t2d0NOCOMPn # aka /dev/rmt/2mn startuse10 enduse 19 statfile/opt/amanda/etc/daily2/tape-slot tapestatus /opt/amanda/etc/daily2/tape-status cleanfile /opt/amanda/etc/daily2/tape-clean config 2 drivenum2 dev /dev/rmt/c5t3d0NOCOMPn # aka /dev/rmt/3mn startuse20 enduse 29 statfile/opt/amanda/etc/daily3/tape-slot tapestatus /opt/amanda/etc/daily3/tape-status cleanfile /opt/amanda/etc/daily3/tape-clean config 3 drivenum3 dev /dev/rmt/c5t4d0NOCOMPn # aka /dev/rmt/4mn startuse30 enduse 47 statfile/opt/amanda/etc/daily4/tape-slot tapestatus /opt/amanda/etc/daily4/tape-status cleanfile /opt/amanda/etc/daily4/tape-clean # dumptypes # # These are referred to by the disklist file. The dumptype specifies # certain parameters for dumping including: # auth- authentication scheme to use between server and client. # Valid values are "bsd" and "krb4". Default: [auth bsd] # comment - just a comment string # comprate- set default compression rate. Should be followed by one or # two numbers, optionally separated by a comma. The 1st is #