When running in Autosync, syncing takes forever. Proposal to autosync selective only out of sync items

[peterbosalliandercom]

Summary

We need to speed up syncing, sometimes it takes hours before syncing is ready. We found that a 'normal' sync applies to all items, also the synced items. This takes much time. When we do a selective sync for all 'out-of-sync' items it is ready in seconds. We like to have an option in the automatic sync to filter all out of sync items. The same as in the UI like here:

Motivation

We cannot use the autosync right now because the apps stay in syncing state.

Proposal

extend the autosync with an option to use selective sync: 'out-of-sync'

[ghostsquad]

I'm curious to know why the syncing takes so long. What does CPU/Memory usage of ArgoCD look like during a sync?

[peterbosalliandercom]

We have a lot of resources sub helm charts etc

[peterbosalliandercom]

We use argo to deploy multiple namespaces with resources in one time.

[kshamajain99]

Looking into this.

[Smirl]

Hey @kshamajain99, Any update on this? Any guidance on an open source PR? Our syncs are taking hours to complete now when we only add/update/remove a couple of objects per commit. In our case we have a chart that deploys tenant namespaces, RBAC, resource quotas, etc. as mentioned in #4521

[ghostsquad]

It would nice to have some more debug information on this, like what is it doing that causes it to take so long? (this is a message to the maintainers). Would debug level logging be enough to get this information?

[kshamajain99]

@Smirl I am currently working on this issue, you can expect an PR soon in a day or two.

[alexmt]

Implemented. Thank you @kshamajain99 !

[peterbosalliandercom]

@kshamajain99 Is this option influencing the post sync hooks?

[Smirl]

Do you have a timeline/milestone for when this will be release? I don't see it in the 1.9 milestone. It would save hours of each sync for us when it is out.

[kshamajain99]

@Smirl This issue is in 1.9 milestone. It will be released in a week or two.