Backup and restore errors

[marmarek]

Reported by marmarek on 23 Dec 2013 15:54 UTC
The new tools needs more testing.

Example of noticed exception during backup restore from AppVM:

AttributeError: 'NoneType' object has no attribute 'read'
at line 213
of file /usr/lib64/python2.7/site-
packages/qubesmanager/restore.py.

----
line: backup_source,vmproc.stderr.read()))
func: restore_vm_dirs
line no.: 882
file: /usr/lib64/python2.7/site-packages/qubes/backup.py
----
line:appvm=appvm)
func: backup_restore_header
line no.: 962
file: /usr/lib64/python2.7/site-packages/qubes/backup.py
----
line: appvm=self.target_appvm)
func: __fill_vms_list__
line no.: 137
file: /usr/lib64/python2.7/site-packages/qubesmanager/restore.py
----
line: self.__fill_vms_list__()
func: current_page_changed
line no.: 213
file: /usr/lib64/python2.7/site-packages/qubesmanager/restore.py

Migrated-From: https://wiki.qubes-os.org/ticket/764

[marmarek]

Modified by joanna on 26 Dec 2013 11:30 UTC

[marmarek]

Modified by joanna on 26 Dec 2013 11:30 UTC

[marmarek]

Modified by joanna on 26 Dec 2013 13:01 UTC

[marmarek]

Modified by joanna on 26 Dec 2013 13:01 UTC

[marmarek]

Comment by marmarek on 29 Dec 2013 02:22 UTC
This particular error was because of wrong path to the backup. And because of bug in error handling this ended up with exception instead of nice message. Fixed here:
http://git.qubes-os.org/?p=marmarek/core-admin.git;a=commit;h=994e963ab88f618b7b4640fbd91101e2bbf7ea93

[marmarek]

Comment by marmarek on 29 Dec 2013 02:22 UTC
This particular error was because of wrong path to the backup. And because of bug in error handling this ended up with exception instead of nice message. Fixed here:
http://git.qubes-os.org/?p=marmarek/core-admin.git;a=commit;h=994e963ab88f618b7b4640fbd91101e2bbf7ea93

[marmarek]

Comment by marmarek on 9 Jan 2014 13:34 UTC
Some mysterious errors:

1. tar2qfile.c - lseek sometimes exits with EAGAIN, which isn't mentioned as possible error of this syscall anywhere. I've added the code to retry 3 times, then fallback to read (ignoring the data).
2. backup.py:780: pipe.close() sometimes fails wit EPIPE. This closes pipe which is input to the final "tar x" (this tar merges multi-volume archive and handles sparseness). EPIPE here means that tar closed its reading end of the pipe, which should mean end of the current volume, but apparently some data left in the output buffer of the pipe (which close() tried to flush). Perhaps some padding or so... I've added the code to ignore this error, in any case we are monitoring exit codes of all the processes involved in backup restore.

Both happened while restoring backup in new format, encrypted, from AppVM, which have device mapper device attached from dom0, where it was USB3 disk. Ok, some diagram:
USB3 disk -> dom0 -> LVM -> xl block-attach -> AppVM
then: AppVM -> tar2qfile -> dom0 -> qfile-unpacker -> openssl decryption -> tar

[marmarek]

Comment by marmarek on 9 Jan 2014 13:34 UTC
Some mysterious errors:

1. tar2qfile.c - lseek sometimes exits with EAGAIN, which isn't mentioned as possible error of this syscall anywhere. I've added the code to retry 3 times, then fallback to read (ignoring the data).
2. backup.py:780: pipe.close() sometimes fails wit EPIPE. This closes pipe which is input to the final "tar x" (this tar merges multi-volume archive and handles sparseness). EPIPE here means that tar closed its reading end of the pipe, which should mean end of the current volume, but apparently some data left in the output buffer of the pipe (which close() tried to flush). Perhaps some padding or so... I've added the code to ignore this error, in any case we are monitoring exit codes of all the processes involved in backup restore.

Both happened while restoring backup in new format, encrypted, from AppVM, which have device mapper device attached from dom0, where it was USB3 disk. Ok, some diagram:
USB3 disk -> dom0 -> LVM -> xl block-attach -> AppVM
then: AppVM -> tar2qfile -> dom0 -> qfile-unpacker -> openssl decryption -> tar

[marmarek]

Comment by marmarek on 15 Jan 2014 04:58 UTC
Many more bugs fixes, but I'll still leave this ticket open and do some more testing.

[marmarek]

Comment by marmarek on 15 Jan 2014 04:58 UTC
Many more bugs fixes, but I'll still leave this ticket open and do some more testing.

[marmarek]

Comment by joanna on 23 Jan 2014 12:44 UTC
When new bugs are discovered we can create a new, specific ticket for this. Closing this one.

[marmarek]

Comment by joanna on 23 Jan 2014 12:44 UTC
When new bugs are discovered we can create a new, specific ticket for this. Closing this one.