|
2018-01-10
§
|
| 15:14 |
<chasemp> |
tools-clushmaster-01:~$ clush -f 1 -w @k8s-worker "sudo puppet agent --enable && sudo puppet agent --test" |
[tools] |
| 15:03 |
<chasemp> |
tools-k8s-master-01:~# for n in `kubectl get nodes | awk '{print $1}' | grep -v -e tools-worker-1001 -e tools-worker-1016 -e tools-worker-1016`; do kubectl cordon $n; done |
[tools] |
| 14:41 |
<chasemp> |
tools-clushmaster-01:~$ clush -w @k8s-worker "sudo puppet agent --disable 'chase rollout'" |
[tools] |
| 14:01 |
<chasemp> |
tools-k8s-master-01:~# kubectl uncordon tools-worker-1001.tools.eqiad.wmflabs |
[tools] |
| 13:57 |
<arturo> |
T184604 cleaned stalled log files that prevented logrotate from working. Triggered a couple of logrorate runs by hand in tools-worker-1020.tools.eqiad.wmflabs |
[tools] |
| 13:46 |
<arturo> |
T184604 aborrero@tools-k8s-master-01:~$ sudo kubectl uncordon tools-worker-1020.tools.eqiad.wmflabs |
[tools] |
| 13:45 |
<arturo> |
T184604 aborrero@tools-worker-1020:/var/log$ sudo mkdir /var/lib/kubelet/pods/bcb36fe1-7d3d-11e7-9b1a-fa163edef48a/volumes |
[tools] |
| 13:26 |
<arturo> |
sudo kubectl drain tools-worker-1020.tools.eqiad.wmflabs |
[tools] |
| 13:22 |
<arturo> |
empty by hand syslog and daemon.log files. They are so big that logrotate won't handle them |
[tools] |
| 13:20 |
<arturo> |
aborrero@tools-worker-1020:~$ sudo service kubelet restart |
[tools] |
| 13:18 |
<arturo> |
aborrero@tools-k8s-master-01:~$ sudo kubectl cordon tools-worker-1020.tools.eqiad.wmflabs for T184604 |
[tools] |
| 13:13 |
<arturo> |
detected low space in tools-worker-1020, big files in /var/log due to kubelet issue. Opened T184604 |
[tools] |
|
2018-01-09
§
|
| 23:21 |
<yuvipanda> |
paws new cluster master is up, re-adding nodes by executing same sequence of commands for upgrading |
[tools] |
| 23:08 |
<yuvipanda> |
turns out the version of k8s we had wasn't recent enough to support easy upgrades, so destroy entire cluster again and install 1.9.1 |
[tools] |
| 23:01 |
<yuvipanda> |
kill paws master and reboot it |
[tools] |
| 22:54 |
<yuvipanda> |
kill all kube-system pods in paws cluster |
[tools] |
| 22:54 |
<yuvipanda> |
kill all PAWS pods |
[tools] |
| 22:53 |
<yuvipanda> |
redo tools-paws-worker-1006 manually, since clush seems to have missed it for some reason |
[tools] |
| 22:49 |
<yuvipanda> |
run clush -w tools-paws-worker-10[01-20] 'sudo bash /home/yuvipanda/kubeadm-bootstrap/init-worker.bash' to bring paws workers back up again, but as 1.8 |
[tools] |
| 22:48 |
<yuvipanda> |
run 'clush -w tools-paws-worker-10[01-20] 'sudo bash /home/yuvipanda/kubeadm-bootstrap/install-kubeadm.bash'' to setup kubeadm on all paws worker nodes |
[tools] |
| 22:46 |
<yuvipanda> |
reboot all paws-worker nodes |
[tools] |
| 22:46 |
<yuvipanda> |
run clush -w tools-paws-worker-10[01-20] 'sudo bash /home/yuvipanda/kubeadm-bootstrap/remove-worker.bash' to completely destroy the paws k8s cluster |
[tools] |
| 22:46 |
<madhuvishy> |
run clush -w tools-paws-worker-10[01-20] 'sudo bash /home/yuvipanda/kubeadm-bootstrap/remove-worker.bash' to completely destroy the paws k8s cluster |
[tools] |
| 21:17 |
<chasemp> |
...rush@tools-clushmaster-01:~$ clush -f 1 -w @k8s-worker "sudo puppet agent --enable && sudo puppet agent --test" |
[tools] |
| 21:17 |
<chasemp> |
tools-clushmaster-01:~$ clush -f 1 -w @k8s-worker "sudo puppet agent --enable --test" |
[tools] |
| 21:10 |
<chasemp> |
tools-k8s-master-01:~# for n in `kubectl get nodes | awk '{print $1}' | grep -v -e tools-worker-1001 -e tools-worker-1016 -e tools-worker-1028 -e tools-worker-1029 `; do kubectl uncordon $n; done |
[tools] |
| 20:55 |
<chasemp> |
for n in `kubectl get nodes | awk '{print $1}' | grep -v -e tools-worker-1001 -e tools-worker-1016`; do kubectl cordon $n; done |
[tools] |
| 20:51 |
<chasemp> |
kubectl cordon tools-worker-1001.tools.eqiad.wmflabs |
[tools] |
| 20:15 |
<chasemp> |
disable puppet on proxies and k8s workers |
[tools] |
| 19:50 |
<chasemp> |
clush -w @all 'sudo puppet agent --test' |
[tools] |
| 19:42 |
<chasemp> |
reboot tools-worker-1010 |
[tools] |
|
2018-01-05
§
|
| 23:49 |
<madhuvishy> |
Run clush -w @k8s-worker -x tools-worker-1001.tools.eqiad.wmflabs 'sudo service docker restart; sudo service flannel restart; sudo service kubelet restart; sudo service kube-proxy restart' on tools-clushmaster-01 |
[tools] |
| 16:22 |
<andrewbogott> |
moving tools-worker-1027 to labvirt1015 (CPU balancing) |
[tools] |
| 16:01 |
<andrewbogott> |
moving tools-worker-1017 to labvirt1017 (CPU balancing) |
[tools] |
| 15:32 |
<andrewbogott> |
moving tools-exec-1420.tools.eqiad.wmflabs to labvirt1015 (CPU balancing) |
[tools] |
| 15:18 |
<andrewbogott> |
moving tools-exec-1411.tools.eqiad.wmflabs to labvirt1017 (CPU balancing) |
[tools] |
| 15:02 |
<andrewbogott> |
moving tools-exec-1440.tools.eqiad.wmflabs to labvirt1017 (CPU balancing) |
[tools] |
| 14:47 |
<andrewbogott> |
moving tools-webgrid-lighttpd-1421.tools.eqiad.wmflabs to labvirt1017 (CPU balancing) |
[tools] |
| 14:25 |
<andrewbogott> |
moving tools-webgrid-lighttpd-1420.tools.eqiad.wmflabs to labvirt1015 (CPU balancing) |
[tools] |
| 14:05 |
<andrewbogott> |
moving tools-webgrid-lighttpd-1417.tools.eqiad.wmflabs to labvirt1015 (CPU balancing) |
[tools] |
| 13:46 |
<andrewbogott> |
moving tools-webgrid-lighttpd-1419.tools.eqiad.wmflabs to labvirt1017 (CPU balancing) |
[tools] |
| 05:33 |
<andrewbogott> |
migrating tools-worker-1012 to labvirt1017 (CPU load balancing) |
[tools] |