[ovirt-devel] Ensure processes death by terminating decorator - https://gerrit.ovirt.org/51407

[Nir Soffer]

On Sat, Jan 23, 2016 at 7:11 PM, Yaniv Bronheim <ybronhei at redhat.com> wrote:
> any updates about that?
>
> https://gerrit.ovirt.org/#/c/52357/ - this can be verified and get it

Should wait for execCmd (it should use the new terminating() context manager
to ensure process is killed on errors).

> https://gerrit.ovirt.org/#/c/52349/ - got tiny comment about unneeded kill
> call

Not ready yet

> https://gerrit.ovirt.org/51407 - Nir can merge

Waiting for Piotr to ack the tests (https://gerrit.ovirt.org/52362)

>
> Nir - please review https://gerrit.ovirt.org/52646 or take over

Should wait for execCmd

>
> and update soon what the plans regarding the async usages in
> vdsm/storage/mount.py
> vdsm/storage/iscsiadm.py
> vdsm/storage/imageSharing.py
> vdsm/storage/hba.py
> vdsm/storage/blockSD.py

I will not have time for this at least after devconf, and even after
devconf we are
busy with spm removal (probably not for 4.0) and ovirt-image (for 4.0).

I will add this to the storage todo list for now.

> and v2v.py

v2v need to read only from stdout. If v2v is not expected to write
more then 64k
errors to stderr (highly unlikely), we don't need to use AsyncProc.
Can create the
process like we do in qemuimg.py, and read from the process stdout.

If we want to handle the unlikely case of huge error output, we can
use CommandStream
to collect the output from both streams, but we will have to the
change the output
parser to be driven by output callback, instead of reading lines from stdout.

> I prefer not to wait for that too long - we can remove the deathSignal
> usages there, and continue with https://gerrit.ovirt.org/#/c/48384

I think we should continue with other Python 3 porting efforts until
we can eliminate
cpopen non-standard apis.

> Please also check if you can take over the re-implementation of async proc
> (https://gerrit.ovirt.org/49441) as you (storage operations) are the main
> and only user of it, and it should fit Popen proc.

I think the current patch does not need too much work - only implement wait
with a timeout.

What about the subproces32 project you found? it looks like the right
direction:
https://pypi.python.org/pypi/subprocess32/

The author is Python core developer:
https://github.com/python/cpython/commits?author=gpshead

We can drop cpopen, AsyncProc, AsyncProcOperation (used only in iscsi), and
use this module to get Python 3 features and reliability, and be compatible with
both Python 3 (fedora 24?) and 2 (el 7.x).

> On Mon, Jan 18, 2016 at 3:12 PM, Francesco Romani <fromani at redhat.com>
> wrote:
>>
>> ----- Original Message -----
>> > From: "Yaniv Bronheim" <ybronhei at redhat.com>
>> > To: "devel" <devel at ovirt.org>, "Shahar Havivi" <shavivi at redhat.com>,
>> > "Francesco Romani" <fromani at redhat.com>, "Nir
>> > Soffer" <nsoffer at redhat.com>
>> > Sent: Monday, January 18, 2016 11:01:10 AM
>> > Subject: Ensure processes death by terminating decorator -
>> > https://gerrit.ovirt.org/51407
>> >
>> > Hi guys,
>> >
>> > Following the work to omit deathSignal attribute from our cpopen
>> > implementation we posted https://gerrit.ovirt.org/51407 which is ready
>> > for
>> > use.
>> > Currently locations that should use it are:
>> > (I wrote above who I expect to check the area and post a patch for that
>> > -
>> > we'll discuss it during next vdsm-sync to follow the work)
>>
>> > fromani:
>> > vdsm_hooks/checkimages/before_vm_start.py - in checkImage - the code
>> > looks
>> > ok, but check if not better to use the terminating decorator.. I think
>> > it
>> > will be nicer
>>
>> Fair enough, posted https://gerrit.ovirt.org/52349
>>
>> > some places define deathSignal for no reason, the call is sync - please
>> > remove those places:
>> [...]
>> > fromani:
>> > lib/vdsm/virtsparsify.py
>>
>> Done in https://gerrit.ovirt.org/52357
>>
>>
>>
>> --
>> Francesco Romani
>> RedHat Engineering Virtualization R & D
>> Phone: 8261328
>> IRC: fromani
>
>
>
>
> --
> Yaniv Bronhaim.