Especially the kube head and the most senior GCE in the pool:
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
svc/kubernetes 10.67.240.1 <none> 443/TCP 26d
26 days of up time! That is impressive, but on the other hand as they stick longer their fluent instance is more likely to fail, it seems (anectodal, no definitive data). And there is also this:
(and do keep in mind that I had to keep restarting my rcc{0-5} DCs almost twice a week. Maybe we can set a regular schedule to reset (down && up) our clusters to make them more error prone. As I was reading the kube documentation I saw a relevant optioned being mentioned here:
https://cloud.google.com/container-engine/docs/node-auto-repair
And as we discussed earlier, every single new kube feature adaption comes with its own problems, but if those old nodes keep getting stuck (see #70) :man_shrugging:
Especially the kube head and the most senior GCE in the pool:
26 days of up time! That is impressive, but on the other hand as they stick longer their fluent instance is more likely to fail, it seems (anectodal, no definitive data). And there is also this:![](https://cl.ly/163z3C1G1W0F/[5b82ccedf41e682d6c7bda8e0ac9d9ef]_Image%202017-06-20%20at%204.01.37%20AM.png)
(and do keep in mind that I had to keep restarting my
rcc{0-5}
DCs almost twice a week. Maybe we can set a regular schedule to reset (down
&&up
) our clusters to make them more error prone. As I was reading the kube documentation I saw a relevant optioned being mentioned here: https://cloud.google.com/container-engine/docs/node-auto-repairAnd as we discussed earlier, every single new kube feature adaption comes with its own problems, but if those old nodes keep getting stuck (see #70) :man_shrugging: