[s3ql] Is it really safe to interrupt upgrade with Ctrl-C?

[Andrei]

Hi,

I was in the process of upgrading from revision 23 to revision 24 and
decided to interrupt it (with Ctrl-C) and resume at a later time.

I got:

"^CERROR: Uncaught top-level exception:
Traceback (most recent call last):
..."

Shouldn't the interrupt have been caught in the code? How do I know I
didn't interrupt a critical operation and corrupted my filesystem?

Thanks

[Daniel Jagszent]

Hi Andrei,

Post by Andrei
[...]
I was in the process of upgrading from revision 23 to revision 24 and
decided to interrupt it (with Ctrl-C) and resume at a later time.
[...]

That should be ok, since the upgrade process checks each metadata if it
is in the old format and only then updates it. Re-running the upgrade
process just re-checks everything from the beginning.

Post by Andrei
Shouldn't the interrupt have been caught in the code? How do I know I
didn't interrupt a critical operation and corrupted my filesystem?

Its unlikely that your filesystem is corrupted since S3QL tries to
minimize the possibility for that. Anyways, after you have successfully
upgraded the filesystem you can check with s3ql_verify whether all
metadata is ok.

If it is not OK you can use the attached python script to re-run the
metadata-upgrade process on an otherwise already upgraded filesystem.
I needed that script since my backend didn't like this many metadata
updates in such a short time or had a bad replication day and I ended up
with an upgraded filesystem with some metadata being in the old format.

[Nikolaus Rath]

Post by Daniel Jagszent
If it is not OK you can use the attached python script to re-run the
metadata-upgrade process on an otherwise already upgraded filesystem.
I needed that script since my backend didn't like this many metadata
updates in such a short time or had a bad replication day and I ended
up with an upgraded filesystem with some metadata being in the old
format.

Out of curiosity: what backend is that? That seems like a terrible
failure mode.

Best,
-Nikolaus

[Nikolaus Rath]

Post by Andrei
Hi,
I was in the process of upgrading from revision 23 to revision 24 and
decided to interrupt it (with Ctrl-C) and resume at a later time.
..."
Shouldn't the interrupt have been caught in the code?

It was :-). I guess what you meant is: should there have been no
stacktrace. I don't have strong opinions about that, but it's extra
work..

Post by Andrei
How do I know I didn't interrupt a critical operation and corrupted my
filesystem?

Because the help text told you so :-)

"You may interrupt the update with Ctrl+C and resume at a later time,
but the filesystem will not be in a usable state until the upgrade is
complete."

Best,
-Nikolaus