Display errors on queued computers (Unicast)

[Tom Elliott]

@tian Can you update and try again? I re-adjusted how the counting is to work, in hopes it will work a little better.

[Cpasjuste]

Hi,

Just a little message to say that i do have the same problem with SVN 6975 : i have an “attempting to check in” loop when trying to restore an image. It was working fine before i updated (i don’t remember which version i was previously but it was SVN 69xx).

[Tom Elliott]

@Cpasjuste What do you mean? Pictures please if possible.

[Cpasjuste]

@Tom-Elliott sure, here it is ! So i create a basic task (deploy), then it just loop on the “attempting to check in”. This task did work fine (same machine/image) a few hours ago just before i update to latest git (svn).

[Cpasjuste]

Just a little more information i just found : if i click the “force task to start” button in the “Task Management”, then the deploy task does start. By the way, if you need some testing just send me a mail. I do have a good linux and developement knowledge which may help to debug if needed.

[Tom Elliott]

@Cpasjuste Should be fixed, sorry about that.

[Cpasjuste]

@Tom-Elliott

Thanks and no problem, we know what we sign for when we use the devel branch I’ll update in a few minutes and let you know.
Edit: i confirm that it’s fixed.

[tian]

@Tom-Elliott Thanks for your hard work. Now the numbers are changing and are different. But these are totally mixed up now (Version 6977):


The pictures were taken from different computers.
A lot of times there still are the same numbers to displayed on different computers at the same time (maybe the 5 seconds interval is too huge to see if there are moments with double numbers).

Is there anything else I can provide?

(In fog 0.32 the queued numbers were/are fine - with the difference that there was/is a delay till free slots can be used by queued computers. Now in the recent versions the computers starting up first gets the free slots - but that is ok.)

[Tom Elliott]

@tian While I’m glad we’re closer, I hope to have finally gotten this more properly solved. The checkin process was constantly updating the time which is why you would see the numbers change (depending on the other host checking in).

It should be good now, Hopefully.

[tian]

@Tom-Elliott said:

@tian While I’m glad we’re closer, I hope to have finally gotten this more properly solved. The checkin process was constantly updating the time which is why you would see the numbers change (depending on the other host checking in).

It should be good now, Hopefully.

I don’t want to tell you - but now (Version 6981) all queued computers display “There are 0 before me” again …(It also would be fine just to display the total number of computers waiting - or just display the time waited - if this problem consumes too much time at the moment…)

[Tom Elliott]

@tian I’m fairly sure it should work properly. I only say this because I have the time setup now and I tested with a 1 client node and 3 tasked hosts.

I don’t know the exact reasoning that these things occur, but I don’t think they’re hindering anything. Maybe group tasks will suck for this?

[tian]

@Tom-Elliott Maybe the group deployment task is not working correctly with this in the trunk versions. Except the wrong display the other things are working. We also just use one server (and on this one the default storage/node). We mostly use group deployment tasks - single deployment tasks are not used that often and wouldn’t reach the limit we set.

Thanks for all the effort you put into solving this.

[Tom Elliott]

@tian My worry for the display is that all the clients have the same time starting out. Because of this, they all get updated at the same time too. I update the check in times if they have timed out now, so order should be maintained. However, if all times are identical, there’s no real way to know which one is first or not. Again, it’s a minimal problem, though I can understand the “huh?” issue.

[Wayne Workman]

@Tom-Elliott But the count issue is solved? The “There are 5 before me” message is accurate?

[Tom Elliott]

@Wayne-Workman The count issue is based off of timing now. So if 5 hosts checkin at the same time, they all have the same count value.
Normally there is a delay, but in the case of group tasks, this might not exist.

Normal cases will be accurate, and I suppose this one is accurate as well.

[Wayne Workman]

@Tom-Elliott Will the number be just all over the place or will it be consistent? Like if 30 computers all check in at exactly the same moment and maximum clients is set to 10 and 10 start, will all other 20 show that “There are 20 before me” ?

Just trying to get an understanding.

[Tom Elliott]

@Wayne-Workman THey base on the time. If all of them have the exact same checkin time, they will ALL say 0 before me, because it doesn’t know which is which, (of which I’ll likely fall back to the ID of the tasking).

For now it’s working but a bit odd.