Opened 6 years ago

Closed 6 years ago

#661 closed defect (invalid)

Mondorestore throws "bzip2: Data integrity error when decompressing."

Reported by: dvedadt9 Owned by: bruno
Priority: normal Milestone: 3.0.4
Component: mondo Version: 3.0.2
Severity: normal Keywords:
Cc:

Description

From time to time we experience errors while perfroming mondorestore "bzip2: Data integrity error when decompressing." That has been confirmed on various servers that are backed up with mondo. Files that were affected with this error were not been changed during backup. Log files attached.

Attachments (1)

mondologs.zip (156.3 KB) - added by dvedadt9 6 years ago.

Download all attachments as: .zip

Change History (3)

Changed 6 years ago by dvedadt9

comment:1 Changed 6 years ago by bruno

  • Milestone changed from 3.0.3 to 3.0.4

1/ don't use -I in your backup script 2/ Use option -V to verify the backup 3/ do you transfer the ISO berfore restoring ? burn them ? check the md5sum of each image before doing these operations. 4/ There are some errors in your archive log file: /home/oracle/app/oracle/agent12g/agent_inst/sysman/emd/upload/pending/D_P4_S0_I586304.xml": No such file or directory Any relationship with your issue ? 5/ try reducing the size of your media, in case the end is causing issue (4400m, or 4300m) 6/ change compression tool (gzip and lzo are also supported) 7/ and check your tarbet HDD to be sure it is sane. 8/ exclude the large files causing issue and backup them separately.

My guess is #3

In the last 10 years I've been dealing with MondoRescue?, I've never seen a previous report linked to an issue with bzip2 per se.

comment:2 Changed 6 years ago by bruno

  • Resolution set to invalid
  • Status changed from new to closed

Received answer by mail: We are on some trail. As backups that we are performing are stored at remote CIFS drive, we found a lot of errors during mondobackup in /var/log/messages like this:

Jan 25 10:01:14 db3 kernel:  CIFS VFS: No response to cmd 47 mid 47801
Jan 25 10:01:14 db3 kernel:  CIFS VFS: Write2 ret -11, wrote 0
Jan 25 10:01:20 db3 kernel:  CIFS VFS: Write2 ret -11, wrote 0
Jan 25 10:01:49 db3 kernel:  CIFS VFS: No response to cmd 47 mid 61296
Jan 25 10:01:49 db3 kernel:  CIFS VFS: Write2 ret -11, wrote 0
Jan 25 10:01:55 db3 kernel:  CIFS VFS: Write2 ret -11, wrote 0
Jan 25 10:02:25 db3 kernel:  CIFS VFS: No response to cmd 47 mid 9384
Jan 25 10:02:25 db3 kernel:  CIFS VFS: Write2 ret -11, wrote 0

It seems that copying files to CIFS share that is mounted without "hard" mount switch have nasty habit that copied file (in case of network problems or congestion) is corrupted. "Hard" switch might help to avoid false feeling that everything is OK, but it does not solve errors during copying of file to CIFS share.

So, mondorescue works as expected, and it seems this is not mondo bug. Problem is with network interconnect ant that is for sure out of mondo's scope.

Note: See TracTickets for help on using tickets.