tree e79db57c447310ff0a359db892fc4aa5f4e19241
parent b3128e0bca7877c3c423d5b0d639803e7f837b13
author mynameborat <bharath.kumarasubramanian@gmail.com> 1615561647 -0800
committer GitHub <noreply@github.com> 1615561647 -0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJgS4OvCRBK7hj4Ov3rIwAAdHIIABCkadUPsJk7OQQYbCbkVgPD
 d2ogBWlijkcZo7NNKEZSqQWQ8B5sMkkchg4Z/P+wwKxrIn1Bop32ONm0En2aSrkt
 sSv4o5mBykBUDH1MZEaHTuBo2wXULNRZhTPkQpD+cS3l2gTVba1AP1FYiMqwH1Lb
 MMPs/SqwUb5sRODVLE9UQTwqvTRryDo2zqG3VKdIiUP8cOhcHoUlvlWqFniGw/jk
 C6eI8cyhr1D4SXnX8+6J24KkTZBcn1+euJEW0O4uSUAebOzJelnFijh1c1lywFpI
 TrmSVVmT7n0VTbD0/YwjlNMD9QS1Zbvy1jevgVn2J0N/BvcqfF+2DwgrssUlByU=
 =CiWB
 -----END PGP SIGNATURE-----
 

SAMZA-2632: Processors should only write their own task locality to zookeeper (#1473)

Problem:
Processors update the task locality for all the tasks in the job model with their own locality regardless of whether the tasks belong to their container model or not.
Description:
As part of SEP 11, host affinity for standalone was introduced. For this feature, we persist the task locality in zookeeper so that subsequent rebalances take this locality into account when generating job model. During job model consensus, we update the task locality for all the tasks and this results in incorrect locality depending on the order of writes.
Changes:
Only update the locality for the tasks that belong to the processor.