sdkserver: When waitForConnection fails, container should restart quickly #3157
Conversation
…ckly * waitForConnection is using the context provided by NewSDKServerContext, but that context isn't closed until we see a Shutdown(). That's not right - we want to restart the SDK server quickly, which might save the game server in this odd case. So, move waitForConnection up into main so that we can use the "normal" context. * Additionally, I noticed that when we went into graceful termination, the liveness probes stopped! Agghhh. So, this reverts part of googleforgames#3072: instead of using /gshealthz as our heartbeat, fall back to using a go routine. But we still honor the intent of googleforgames#3072: health checks should start when we see the first /gshealthz. * Along the way, add some logging, and make it consistently "SDK" in logging.
zmerlynn
requested review from
gongmax and
roberthbailey
and removed request for
markmandel and
EricFortin
|
Build Failed 😱 Build Id: 39818ae3-b6ca-4bfb-bafd-269e5ed73cde To get permission to view the Cloud Build view, join the agones-discuss Google Group. |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: gongmax, zmerlynn The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
zmerlynn
force-pushed
the
rework-sdkserver-health
branch
from
462d31a
to
f15d973
Compare
|
New changes are detected. LGTM label has been removed. |
|
Build Failed 😱 Build Id: 09ff8d9f-8fe8-42b8-96a1-652795491fbc To get permission to view the Cloud Build view, join the agones-discuss Google Group. |
|
Build Succeeded 👏 Build Id: 16ff04af-942c-4a6a-b8d7-7aefcb12ef15 The following development artifacts have been built, and will exist for the next 30 days:
A preview of the website (the last 30 builds are retained): To install this version:
|
|
Build Failed 😱 Build Id: e31c2c53-8a00-4b71-80a8-8e6251810340 To get permission to view the Cloud Build view, join the agones-discuss Google Group. |
|
Build Succeeded 👏 Build Id: 98d7574e-f8cc-4773-9cda-e52606b8d5ce The following development artifacts have been built, and will exist for the next 30 days:
A preview of the website (the last 30 builds are retained): To install this version:
|
waitForConnection is using the context provided by NewSDKServerContext, but that context isn't closed until we see a Shutdown(). That's not right - we want to restart the SDK server quickly, which might save the game server in this odd case. So, move waitForConnection up into main so that we can use the "normal" context.
Additionally, I noticed that when we went into graceful termination, the liveness probes stopped! Agghhh. So, this reverts part of Rework game server health initial delay handling #3072: instead of using /gshealthz as our heartbeat, fall back to using a go routine. But we still honor the intent of Rework game server health initial delay handling #3072: health checks should start when we see the first /gshealthz.
Along the way, add some logging, and make it consistently "SDK" in logging.