Re: [Qemu-devel] [PATCH v4 2/2] qemu-ga: sample fsfreeze hooks

[Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com>]

On 2012/11/26 21:40, Luiz Capitulino wrote:
> On Mon, 26 Nov 2012 20:49:17 +0900
> Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com> wrote:
> 
>> On 2012/11/23 1:03, Luiz Capitulino wrote:
>>> On Thu, 22 Nov 2012 11:15:49 +0900
>>> Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com> wrote:
>>>
>>>> Adds sample hook scripts for --fsfreeze-hook option of qemu-ga.
>>>>   - fsfreeze-hook : execute scripts in fsfreeze-hook.d/
>>>>   - fsfreeze-hook.d.sample/mysql-flush.sh : quiesce MySQL before snapshot
>> <snip>
>>>> +# Iterate executables in directory "fsfreeze-hook.d" with the specified args
>>>> +[ ! -d "$FSFREEZE_D" ] && exit 1
>>>> +for file in "$FSFREEZE_D"/* ; do
>>>> +    is_ignored_file "$file" && continue
>>>> +    [ -x "$file" ] || continue
>>>> +    echo "$(date): execute $file $@" >>$LOGFILE
>>>> +    "$file" "$@" >>$LOGFILE 2>&1
>>>> +    STATUS=$?
>>>> +    echo "$(date): $file finished with status=$STATUS" >>$LOGFILE
>>>> +done
>>>
>>> execute_fsfreeze_hook() will fail the freeze process if this script fails. Two
>>> comments:
>>>
>>>  1. Do we want to fail the freeze process if one of the sub-scripts fails?
>>>     If yes, then we have to exit 1 in the first failure
>>
>> I originally thought the hooks are optional; even they failed, filesystem-level
>> consistency are still kept in the snapshot.
>> However, if we are going to fail fsfreeze process by one of sub-scripts'
>> failure, we also need to notify scripts which already succeeded to pre-freeze
>> to thaw (or abort) freezing, before exit 1 here.
> 
> Right, which makes things more complex. I vote for doing it the simpler way
> for now then, which is to ignore subscripts exit status. But then you should
> add exit 0 after the loop to avoid this:

I agree this way.

>>>  2. The exit status of the script will echo's exit status. I doubt we want that
>>
>> Do you mean $STATUS (not $status) is specialized in some shell environments?
> 
> No, what I meant is that (afaik) when the script finishes, its return status to
> qemu-ga is actually going to be the latest call echo exit status because it's
> the last command executed in the loop. If echo fails (say no space) then
> the script will fail.
> 
> We either, ignore any failures in qemu-ga itself (although we should at least
> print a warning there) and/or add exit 0 as the last line of the script.

Now I got the point. I will fix this at the next version.

Thanks,
-- 
Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com>
Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory