On Thu, Sep 20, 2018 at 5:08 PM Piotr Kliczewski <pkliczew(a)redhat.com>
wrote:
On Thu, Sep 20, 2018 at 12:55 AM, Germano Veit Michel <germano(a)redhat.com>
wrote:
>
>
> On Thu, Sep 20, 2018 at 7:14 AM Nir Soffer <nsoffer(a)redhat.com> wrote:
>
>>
>>
>> On Mon, 17 Sep 2018, 8:06 Germano Veit Michel, <germano(a)redhat.com>
>> wrote:
>>
>>> Hello,
>>>
>>> I've been struggling with TaskID/FlowID when talking to the VDSM API. I
>>> am trying to write a tool that uses the vdsm-api to facilitate the
>>> troubleshooting of image issues (snapshots). This tool does a series of API
>>> calls, but I cannot find a nice way to track the taskID and clear the
>>> completed tasks after completion of the tool commands. Currently I'm
>>> clearing all Tasks that match the verb and are finished, which is not
>>> ideal. I would like to have the exact TaskID to track and dont want to
>>> leave Tasks behind. I don't want also to clear tasks from other entities
>>> (like engine!).
>>>
>>
>> Storage jobs are managed using (client generated) job id. The task id is
>> internal
>> implementation detail which will likely disappear in future version. You
>> don'need
>> to monitor or clean the internal tasks, they are managed by the storage
>> jobs
>> framework for you.
>>
>
> Thanks for the clarification Nir!
>
>
>> You can check engine code to understand how storage jobs are managed.
>>
>>
>>> I understand that if I want to specify the task/flow ids when calling
>>> the vdsm API, these two need to be passed as headers (http) so they end up
>>> in the context of the call. Is this correct? But using vdsm/client.py I
>>> cannot figure out how to do this, but I understand it is possible.
>>>
>>
>> This is not possible now, but sounds like a good idea. Please file RFE
>> to add this,
>> or if you have the time you can try to add. Looks like the place you can
>> add it is
>> in lib/yajsonrpc/stompclient.py - ClientRpcTransportAdapter.send().
>>
>
> Nice, I'll take a look.
>
>
>>
>> Piotr, what do you think?
>>
>
I agree. In situations like this it would be great to be able to provide
flow_id. It is possible to provide it in java based client [1].
Please open an RFE and we will implement it for you.
Thanks! I can help testing it as soon as you submit a patch, will cherry
pick it once you send.
RFE:
https://bugzilla.redhat.com/show_bug.cgi?id=1631587