Remove unused features (socket activation and KSM) #177
Conversation
|
Note that this cannot realistically land until we a stand-alone daemon to handle KSM, hence the dnm label. See: |
|
kubernetes qa-failed 👎 |
|
The tests are failing due to https://github.com/clearcontainers/tests/blob/master/.ci/install_proxy.sh#L34..L35 which is trying to modify the proxy's service file (which has been removed by this PR). Once we have a KSM daemon, we can then:
|
Now that the cc-proxy will not be a service, we need to use option -t in journalctl to get the logs from the proxy. Relates to: clearcontainers/runtime#835 clearcontainers/proxy#177 Fixes: clearcontainers#775 Signed-off-by: Salvador Fuentes <[email protected]>
Now that the cc-proxy will not be a service, we need to use option -t in journalctl to get the logs from the proxy. Relates to: clearcontainers/runtime#835 clearcontainers/proxy#177 Fixes: clearcontainers#775 Signed-off-by: Salvador Fuentes <[email protected]>
|
kubernetes qa-failed 👎 |
The proxy should not (cannot) handle KSM now that an instance is being launched per pod. Partially fixes clearcontainers#176. Signed-off-by: James O. D. Hunt <[email protected]>
Socket activation is no longer required since a proxy instance is launched by the runtime for each pod. Fixes clearcontainers#176. Signed-off-by: James O. D. Hunt <[email protected]>
jodh-intel
force-pushed
the
remove-unused-features
branch
from
7e120ec to
e3ab55e
Compare
|
kubernetes qa-failed 👎 |
1 similar comment
|
kubernetes qa-failed 👎 |
|
Blocked on clearcontainers/tests#913. We cannot remove the KSM functionality from the proxy until the issue above is resolved since the tests currently install the proxy "manually" using: (Note that user systems would not be affected by removing the functionality since they have the |
|
Hi @chavafg, @grahamwhaley - now that clearcontainers/tests#920 has landed, could you both confirm that I can continue with removing these features from the proxy? |
|
Let me check some stuff. afaict when we landed clearcontainers/tests#911 then the metrics CI got to do the right thing, but it then appears when we landed clearcontainers/tests#920 it might be back to being nobbled again (that is, KSM seems to be in effect even when we try to turn it off). (oh, and apparently you need a rebase I think any how ;-) ) |
|
yep - I'll hold off rebasing until I get two confirmations :) |
No description provided.