print "emergency deploy to docker"

Question

print "emergency deploy to docker"

danduk82 opened this issue 7 years ago · 12 comments

As discussed with @hpchrist instead of fixing the undebuggable problem we have with tomcat on prod, we are pushing the print tomcat backend in rancher directly.

@procrastinatio : use this address for the internal print ELB for DEV:
vpc-lb-print-dev.intra.bgdi.ch

I have configured the port 8009 on the ELB.

Andrea

Answer 1 · 2017-11-08T13:20:29.000Z

the machines for dev are running and registered in rancher with the labels
application=print and staging=dev

Answer 2 · 2017-11-08T13:20:45.000Z

The elb and DNS are ready

vpc-lb-print-dev.intra.bgdi.ch:8009

Answer 3 · 2017-11-08T13:49:45.000Z

Health check
GET /service-print-main/pdf/info.json HTTP/1.1

Answer 4 · 2017-11-08T14:26:40.000Z

for the print we should mount:
/var/local/efs-applications-rw/print/${staging} into /var/local/print of the containers

Answer 5 · 2017-11-09T07:43:45.000Z

OK, it works on dev. My war did miss the variable specifying the location of the tempDir as /var/local/print

Answer 6 · 2017-11-09T10:08:06.000Z

mmh, I cannot understand why but the new machines do not want to mount the EFS... It's broken on dev too right now...

Answer 7 · 2017-11-09T12:10:14.000Z

for the records, the precedent problems was related to the fact that I was trying to use m3 instances instead of m4. There might be some nasty mount things on boot of m3 types with ephemeral storage (probably).

Answer 8 · 2017-11-09T12:12:56.000Z

The production cluster is ready and registered in rancher.

The puppet recipes for the mf0-legacy machines are working. I have deployed the new EFS only on the the 2 prod machines that are down, so we will have to deploy the new version on them, remove manually the other one, check that it works. If this work, we can deploy on the last one and we are done.

Please tell me when you have re-tested the makefile for the dev and int lines before doing it on prod.

Answer 9 · 2017-11-10T09:29:11.000Z

The production cluster is ready and registered in rancher.

The puppet recipes for the mf0-legacy machines are working. I have deployed the new EFS only on the the 2 prod machines that are down, so we will have to deploy the new version on them, remove manually the other one, check that it works. If this work, we can deploy on the last one and we are done.

Please tell me when you have re-tested the makefile for the dev and int lines before doing it on prod.

Answer 10 · 2017-11-10T09:30:30.000Z

@procrastinatio
The last machine is ready for deploy (new EFS mount). Could you please take care of deploying the service on it? I will add it again in varnish after that.

Andrea

Answer 11 · 2017-11-10T11:00:30.000Z

@danduk82 OK, done. Tested info.jsonthrough apache and ELB.

Answer 12 · 2017-11-10T11:05:24.000Z

@procrastinatio ,
I have reactivated the routing on that machine too.