The datahub helm chart makes Frontend and GMS as s...
# all-things-deploymentb
better-orange-49102
03/24/2022, 5:45 AMThe datahub helm chart makes Frontend and GMS as separate deployments
im wondering if anyone has tried to colocate both containers in the same pod, ie, FE as the primary container and GMS as a sidecar?
I'm thinking of trying it out because occasionally, I get people complaining they can't login, and the logs in frontend suggest DNS lookup isn't successful at that point in time. And its a randomly happening issue.
better-orange-49102
03/24/2022, 5:50 AMthe logs in frontend whenever the login has a problem looks like:
Copy code
Just-in-time provisioning in progress....
Attempting to extract use from OIDC profile {redacted}
[application-akka.actordefault-dispatcher-193] Attempting to provision user with urn urn:li:corpuser:{redacted}
[R2 Netty Scheduler 2-1] DEBUG c.l.r.t.h.c ExponentialBackOffRateLimiter - Running rate limited task at 1647392715871 with period 1600
[R2 Nio event Loop-1-8] DEBUG c.l.r.t.h.c ExponentialBackOffRateLimiter - minimum period changed from 1600 to 3200
[R2 Nio event Loop-1-8] DEBUG c.l.r.t.http.client.AsyncPoolImpl - datahub-gms/<some ip address>/<some random string>: object creation failed
com.linkedin.r2.RetriableRequestException: io.netty.channel.AbstractChannel$AnnotationConnectException: Connection refused: datahub-gms/<some ip address>:8080
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection redsed: datahub-gms/<some ip address>:8080
Caused by: java.net.ConnectException: Connection refusede
early-lamp-41924
03/28/2022, 7:33 PMSorry about the late message. Was datahub-gms alive at this time point? Putting them in the same pod doesn’t seem like the good long term solution here. Wondering if gms got too many requests at this time
b
better-orange-49102
03/28/2022, 10:34 PMIt was only 2 people logging in at that time and both encountered the error (and they told me about it) On the other hand, I was able to navigate in datahub at the same time for my logged in session without issue. (I should have logged out and try logging in back then)
better-orange-49102
03/28/2022, 11:48 PMFrom a deployment standpoint, do you think there is any issue with spinning up multiple instances of these gms-frontend pods?
e
early-lamp-41924
03/29/2022, 1:07 AMthere shouldn’t be, but there is a known issue with ui ingestion that happens when you deploy multiple mae-consumers
early-lamp-41924
03/29/2022, 1:07 AMso if you are running gms and consumers int he same pod, this may cause issues
b
better-orange-49102
03/29/2022, 1:10 AMif i decouple gms from mae-consumer, would that work? My consumers are running separately from GMS now
e
early-lamp-41924
03/29/2022, 1:14 AMYes!!!
b
better-orange-49102
03/29/2022, 1:16 AM👍 thanks for the advice!
2 Views