Bug 1576817

Engine shows pasue by system:
2018-05-10 12:19:05 -  EVENT_ID: UPLOAD_IMAGE_PAUSED_BY_SYSTEM_TIMEOUT
 
deamon log shows another write(upload) started 2 minutes later at 2018-05-10
12:21:14
2018-05-10 12:21:14,835 INFO    (Thread-5) [web] START [10.35.162.7] PUT
/images/c3db24af-70b2-45e4-9d3d-fca7872cb184
2018-05-10 12:21:14,836 INFO    (Thread-5) [images] Writing 16777216 bytes
at offset 33554432 flush True to

Created attachment 1434350 [details]
engine, vdsm , proxy , deamon logs

If the transfer was paused because of inactivity, it means inactivity_timeout
seconds passed since the last client request was finished.

If the client is using code like:

   while not done:
       send PUT request
       write data
       sleep inactivity_timeout + 10

Then I think we can expect that the ticket will paused, and the next PUT request
will fail with an error like "403 Forbidden", with a json error about 
"Transfer was paused"

We don't want to remove the ticket since we will loose the ticket state, but we
can add a "paused" attribute.

To pause a transfer, engine will ask vdsm to modify the ticket:

    Host.update_image_ticket(ticket_uuid, {"paused"=true})

Vdsm will send this request to the daemon:

    PUT /tickets/ticket-uuid
    ...
    {
        "paused": true,
    }

To resume a ticket, engine will send this to vdsm:

    Host.update_image_ticket(ticket_uuid, {"paused"=false, "timeout"=300)

Vdsm will send this request to the daemon:

    PUT /tickets/ticket-uuid
    ...
    {
        "paused": false,
        "timeout": 300
    }

Host.update_image_ticket() will replace Host.extend_image_ticket()
which will be deprecated and removed in future version.

Daniel, what do you think?

(In reply to Nir Soffer from comment #3)
> If the transfer was paused because of inactivity, it means inactivity_timeout
> seconds passed since the last client request was finished.
> 
> If the client is using code like:
> 
>    while not done:
>        send PUT request
>        write data
>        sleep inactivity_timeout + 10
> 
> Then I think we can expect that the ticket will paused, and the next PUT
> request
> will fail with an error like "403 Forbidden", with a json error about 
> "Transfer was paused"
> 
> We don't want to remove the ticket since we will loose the ticket state, but
> we
> can add a "paused" attribute.
> 
> To pause a transfer, engine will ask vdsm to modify the ticket:
> 
>     Host.update_image_ticket(ticket_uuid, {"paused"=true})
> 
> Vdsm will send this request to the daemon:
> 
>     PUT /tickets/ticket-uuid
>     ...
>     {
>         "paused": true,
>     }
> 
> To resume a ticket, engine will send this to vdsm:
> 
>     Host.update_image_ticket(ticket_uuid, {"paused"=false, "timeout"=300)
> 
> Vdsm will send this request to the daemon:
> 
>     PUT /tickets/ticket-uuid
>     ...
>     {
>         "paused": false,
>         "timeout": 300
>     }
> 
> Host.update_image_ticket() will replace Host.extend_image_ticket()
> which will be deprecated and removed in future version.
> 
> Daniel, what do you think?

We can just add a new api (e.g. pause_image_ticket), which could be simpler. Or do we expect any more values we might want the update in the ticket later on?

(In reply to Daniel Erez from comment #4)
> We can just add a new api (e.g. pause_image_ticket), which could be simpler.
> Or do we expect any more values we might want the update in the ticket later
> on?

I don't know if we will need new ways to update the ticket, but having a generic
API means we will have to change only engine next time, and it means we need 
single update() API instead of pause() and resume().

This bug didn't get any attention for a while, we didn't have the capacity to make any progress. If you deeply care about it or want to work on it please assign/target accordingly

This bug didn't get any attention for a while, we didn't have the capacity to make any progress. If you deeply care about it or want to work on it please assign/target accordingly

ok, closing. Please reopen if still relevant/you want to work on it.

ok, closing. Please reopen if still relevant/you want to work on it.