Maik Aussendorf <maik.aussend...@bareos.com> writes: > Am 08.02.2017 um 20:29 schrieb Kjetil Torgrim Homme: >> so, trying out restore: >> >> 08-Feb 19:28 dump-sd JobId 1457958: Forward spacing Volume >> "incr/Incr-6247" to file:block 1:970713522. >> 08-Feb 19:28 ece026-snapshot-fd JobId 1457958: Error: >> restore.c:1239 Write error on >> /tmp/bacula-restores/_percona/xbstream.0001457954: Child exited with >> code -268435457 >> 08-Feb 19:28 ece026-snapshot-fd JobId 1457958: Error: Restore >> command returned non-zero value: 1, message: > Here after 'message: ' the stdError output of the xbstream is > expected.
>> I tried to grab the error message from xbstream, but was unable to patch >> the python plugin to log it. however strace delivers the goods, sort > You can try to use 'stdOut' instead of 'stdError' in line 347 (obviously > stdErr is empty in you case). I will print both values in future > versions. yes, I don't understand this. the error message from xbstream is clearly being written to stderr, I don't understand why it doesn't work (I changed the plugin code to pass stderr=PIPE to Popen.) >> of. the xbstream process reads and reads and reads, and then: >> >> write(2, "xb_stream_read_chunk(): invalid checksum at offset >> 0xa00025: expected 0xf96cf5a6, read 0xc7ae89da.\n", 99) = 99 >> >> at this point I would like to see if the problem is in the backup or in >> the restore. can I extract the stream to a plain file on the client and >> run xbstream on it manually? other hints on how to debug the problem? > > Yes, you can. E.g. restore to a client with plugins disabled, then you > will get the plain xbstream-file. ah! ok, so I did this. the resulting file has 8 NUL bytes at the beginning which I chopped off, and I get the same error on stderr when running manually: $ xbstream -x -C extr < /tmp/offs xb_stream_read_chunk(): invalid checksum at offset 0xa00025: expected 0xdfbd8ba6, read 0x8866f355. (the checksums are different from the previous mail since I made a new full backup, but the offset is the same.) running xtrabackup manually gives a file which xbstream accepts. it is just 3 bytes smaller than /tmp/offs. the size of the file is 121330646, 115 MiB - the error occurs just after the 10 MiB mark. at the 10 MiB mark there are 41 NUL bytes before there is a new header - this looks the same in the working dump taken manually. ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@ ^@^@^@^@^@^@^@^@^@XBSTCK01^@P^G^@^@^@ibdata1[...] in other words, at offset 0xa00025 there are 4 NUL bytes, then there is the new header. the md5sum's do not match up: from Bareos database: 1b8aee24ef2856fe0dbfe42214f0c383 /_percona/xbstream.0001459278 restored without plugin: f776b9287ab4cf9e349da41b7aa462b4 /tmp/xbstream.0001459278 above file with first 8 NUL bytes removed: d4f5949c8cc8a2914909b918c0e88bdb /tmp/offs > I suggest that you first check, that xtrabackup produces a valid stream > that can be successfully used by xbstream again. If that works you can > have a look about the stream going into Bareos and what comes back on > restore. thank you for your hints! I'm trying to dig further... -- Kjetil T. Homme Redpill Linpro AS - Changing the game -- You received this message because you are subscribed to the Google Groups "bareos-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to bareos-users+unsubscr...@googlegroups.com. To post to this group, send email to bareos-users@googlegroups.com. For more options, visit https://groups.google.com/d/optout.