task(fix-it-fridays): Speeding up k8s reset #1599
Conversation
| @@ -79,15 +79,15 @@ fi | |||
|
|
|||
| if g3k_manifest_lookup .versions.fence 2> /dev/null; then | |||
| # data ecosystem sub-commons may not deploy fence ... | |||
| gen3 kube-setup-fence | |||
| gen3 kube-setup-fence & | |||
There was a problem hiding this comment.
It it safe to fork some of these setup jobs? I think some apps like fence assume that the audit service ran first to set up the SQS queue, and if things run out of order an app could start with missing resources
There was a problem hiding this comment.
That's a good point. There must be inter-dependencies between the apps...
Perhaps the Kubernetes ReadinessProbe should ping the /status endpoint and the status check should make sure the other micro-service is up and running -- then the services could be self-aware and we wouldn't need to orchestrate that at the cloud-automation / infra provisioning stage.
| else | ||
| gen3_log_info "not deploying portal - no manifest entry for .versions.portal" | ||
| fi | ||
|
|
||
| gen3_log_info "enable network policy" | ||
| gen3 kube-setup-networkpolicy "enable" || true | ||
| gen3 kube-setup-networkpolicy "enable" || true & |
There was a problem hiding this comment.
I think you may actually want...
| gen3 kube-setup-networkpolicy "enable" || true & | |
| {gen3 kube-setup-networkpolicy "enable" || true} & |
...if I'm understanding this right. Otherwise I think the first command will block.
There was a problem hiding this comment.
There was no response here.
@williamhaley , do you want to push a commit with your suggestion?
Quick experiment