Conversation
|
I could in theory make this backwards compatible, but I think it's way nicer to just bump the major and require workers and schedulers to be updated at the same time. |
Wondering if we need to bump the major here. This will make schedulers and workers using different versions incompatible but IIRC we've never said that different versions are compatible with each other. We've generally done compatibility bumps only for client config breaks. |
2246ee2 to
6b61f0c
Compare
ce7b882 to
5c1e9dd
Compare
6b61f0c to
f1e1e05
Compare
5c1e9dd to
2d33b5d
Compare
f1e1e05 to
ecd52ac
Compare
2d33b5d to
758c63f
Compare
palfrey
left a comment
There was a problem hiding this comment.
@palfrey reviewed 8 of 8 files at r1, all commit messages.
Reviewable status: 0 of 1 LGTMs obtained, and all files reviewed
palfrey
left a comment
There was a problem hiding this comment.
Reviewable status:
complete! 1 of 1 LGTMs obtained, and all files reviewed
ecd52ac to
ae25509
Compare
758c63f to
77e8307
Compare
|
#1975 is now merged, and I'm guessing that's the source of conflicts here |
77e8307 to
0edaf6e
Compare
|
Rebased which sorted it out. Happy to merge this if you are. |
0edaf6e to
1ce866f
Compare
Seeing build failures now |
1ce866f to
6997213
Compare
e9a793e to
077070b
Compare
077070b to
e116664
Compare
e116664 to
389a5e1
Compare
389a5e1 to
c5d6166
Compare
c5d6166 to
81dc012
Compare
81dc012 to
eb301cd
Compare
When there are multiple schedulers in high availability mode then the workers can end up communicating with the wrong scheduler and state can get confused. Modify the communications such that there is a single bi-directional gRPC stream between the client and the worker. This way they will always be one-to-one mapped even with a load balancer in front.
eb301cd to
ac8f8ad
Compare
When there are multiple schedulers in high availability mode then the workers can end up communicating with the wrong scheduler and state can get confused. Modify the communications such that there is a single bi-directional gRPC stream between the client and the worker. This way they will always be one-to-one mapped even with a load balancer in front. Co-authored-by: Chris Staite <chris@yourdreamnet.co.uk>
Description
When there are multiple schedulers in high availability mode then the workers can end up communicating with the wrong scheduler and state can get confused.
Modify the communications such that there is a single bi-directional gRPC stream between the client and the worker. This way they will always be one-to-one mapped even with a load balancer in front.
Type of change
Please delete options that aren't relevant.
not work as expected)
How Has This Been Tested?
Running on my cluster this now functions perfectly with a one-to-one worker to scheduler and removes the races.
Checklist
bazel test //...passes locallygit amendsee some docsThis change is