This post is more than 5 years old
44 Posts
1
1168
scanner oddities - watch out for continuation save sets!
So it happened, that we accidentally deleted a tape, with file and media index.
Consequently we re-read the tape with the scanner command (scanner -s nsrhost -m -p -z -v ...).
What confused us a lot was the fact, that not all of the SSIDs that re-appeared in the media index were recoverable (or cloneable in our case):
An attemt to clone only reported
...
5767:nsrclone: NSR server : busy
5766:nsrclone: waiting 30 seconds then retrying
...
And at the server, we got the alert
Tue 12:16:09 PM media waiting: Waiting for 1 writable volume(s) to
backup pool ' ' disk(s) or tape(s) on
After a lot of researching we finally came to the insight that we
needed to also scan the "neighbouring" tapes, so that Networker got
the complete picture of savesets that span multiple tapes.
The overwhelming output of the scanner command run on the tape that
was accidentally deleted, had a hint:
...
scanner: CONT NOTE ssid 4236283626, off 518781797512, flags 0x10103, info 1740881526
scanner: ssid 4236283626: continued from ` ', scan for synchronization
...
The flags that the Saveset 4236283626 showed via mminfo were vF (for
"valid" and "finished (ended)") had given us no notion of the fact
that the SSID was continued: we would have expected an additional "C".
What is your opinion: is this a bug?
In any case, the man-page of the scanner command should contain a tip,
that there often will be continuation save sets that might need a
re-scan with scanner, too.
Kind regards, Tom
bingo.1
2.4K Posts
0
August 5th, 2016 05:00
This is not a bug - it is "good old NetWorker".
You should have looked at mminfo's 'sumflags' attribute and you should have seen that. For example:
... ssid fl
12345 cb complete & browsable
23456 hb header & browsable
23456 mb middle & browsable
23456 tb tail & browsable
Be aware that there could be more than one 'middle' part.
So how to check for the correct sequence? - use mminfo again like in this example:
swadm
44 Posts
0
August 5th, 2016 06:00
Thanks a lot for that tip!
I looked at so many columns of mminfo, but did not realize that sumflags would have answered the question about the saveset completeness.
Thanks and regards, Tom