libgearman isn't hashing submitted jobs based on unique

[SpamapS]

In the original gearman client, the client would take the list of servers it knows about, hash the unique ID of each submitted job, and choose one of the servers in the list based on the hash of the unique.

This code seems to have been lost, and no tests enforce it. This is an important part of what makes gearman awesome, so it needs to be resurrected.

[sirgalleto]

@SpamapS I actually have a bug with the queue creation with mysql because the queue is trying to create a duplicate entry in the database. Please tell me if we are talking about the same bug and if you found any kind of solution.

Thanks

[p-alik]

@sirgalleto, see mein comment in the isseu #80

[SpamapS]

Indeed, @sirgalleto this is not related. This is about clients of many gearmand's that are selecting the wrong gearmand to submit their job to.

[p-alik]

In the original gearman client, the client would take the list of servers it knows about, hash the unique ID of each submitted job, and choose one of the servers in the list based on the hash of the unique.

@SpamapS, could you point me to the code part you mentioned here?

[bmeynell]

@p-alik - This issue appears to have been discovered here and first logged here. This is a huuuuuge gaping hole preventing gearmand from being a true "distributed nervous system". I would also be keen to see the portion(s) of the C code that would need modifications if anyone has that on their radar.