quorum

Message boards : Number crunching : quorum
Message board moderation

To post messages, you must log in.

AuthorMessage
Sergey Kovalchuk

Send message
Joined: 24 Jun 20
Posts: 26
Credit: 1,106,925
RAC: 1
Message 79 - Posted: 28 Jun 2020, 3:34:44 UTC

minimum quorum 2
initial replication 2

in case of an error, a new task is generated but is not sent to clients, it waits in the queue of the scheduler

repeat tasks need to be prioritized over new tasks
ID: 79 · Report as offensive
Profile chip
Project administrator

Send message
Joined: 14 Jun 20
Posts: 78
Credit: 1,321,619
RAC: 0
Message 86 - Posted: 28 Jun 2020, 5:50:01 UTC - in response to Message 79.  

minimum quorum 2
initial replication 2

in case of an error, a new task is generated but is not sent to clients, it waits in the queue of the scheduler

repeat tasks need to be prioritized over new tasks


Ah yeah, I agree. Any idea how to implement this?
ID: 86 · Report as offensive
Jord
Volunteer moderator
Help desk expert
Avatar

Send message
Joined: 24 Jun 20
Posts: 85
Credit: 207,156
RAC: 0
Message 91 - Posted: 28 Jun 2020, 14:59:26 UTC - in response to Message 86.  
Last modified: 28 Jun 2020, 14:59:58 UTC

The transitioner should do this automatically and send it back to the feeder. You can probably play with the feeder's commands for this.
ID: 91 · Report as offensive
Profile KPX

Send message
Joined: 26 Jun 20
Posts: 3
Credit: 17,664,576
RAC: 39,251
Message 105 - Posted: 29 Jun 2020, 11:33:09 UTC

I guess this still needs to be addressed - there are still lots of "validation pending" units, where the validation unit has not been sent out.
ID: 105 · Report as offensive
zombie67 [MM]
Avatar

Send message
Joined: 24 Jun 20
Posts: 21
Credit: 181,137,041
RAC: 272,097
Message 107 - Posted: 29 Jun 2020, 13:12:16 UTC

Perhaps the issue is made worse by having 230,356 tasks already in the queue ready to send. With a pool that large, it may take many days for the resends to get issued.
Reno, NV
Team: SETI.USA
ID: 107 · Report as offensive
Profile chip
Project administrator

Send message
Joined: 14 Jun 20
Posts: 78
Credit: 1,321,619
RAC: 0
Message 110 - Posted: 29 Jun 2020, 22:56:07 UTC - in response to Message 107.  

Perhaps the issue is made worse by having 230,356 tasks already in the queue ready to send. With a pool that large, it may take many days for the resends to get issued.


Yeah, I'm still looking into this, I assume there's a solution without just reducing the pool size.
ID: 110 · Report as offensive
Jord
Volunteer moderator
Help desk expert
Avatar

Send message
Joined: 24 Jun 20
Posts: 85
Credit: 207,156
RAC: 0
Message 111 - Posted: 30 Jun 2020, 0:59:29 UTC - in response to Message 110.  

As always, you can email the BOINC developers email list or the BOINC Projects email list (or both). No need to reinvent BOINC or try to do everything alone.
https://boinc.berkeley.edu/trac/wiki/EmailLists
ID: 111 · Report as offensive
Profile chip
Project administrator

Send message
Joined: 14 Jun 20
Posts: 78
Credit: 1,321,619
RAC: 0
Message 119 - Posted: 1 Jul 2020, 13:16:36 UTC

We've taken a few manual steps to reprioritize jobs which were created as a result of a computation error. This will be running as a 3 hourly task:

while read line; do
    if [[ "$(echo "${line}" | awk '{print $2}')" -lt 2 ]]; then
        mysql -D minecrafthome -s -N -e " \
            UPDATE result \
            SET priority = 1 \
            WHERE id = $(echo ${line} | awk '{print $1}')"
    fi
done <<< $(mysql -D minecrafthome -s -N -e " \
    SELECT id, COUNT(workunitid) \
    FROM result \
    WHERE server_state = 2 AND priority = 0 AND workunitid in \
        (SELECT * FROM \
             (SELECT workunitid \
              FROM result \
              GROUP by workunitid \
              HAVING COUNT(workunitid) > 2) \
         AS a) \
    GROUP BY workunitid;")
ID: 119 · Report as offensive

Message boards : Number crunching : quorum