Hack to stop dataloader related errors

[stepanstipl]

This is an ugly hack to stop TF/Helm data loader errors that freezes data loader version.

It should be removed in scope of #163, once the "new" dataloader is ready to be used.

After potentially failing once (if the dataloader is at older version), no more errors should appear, as TF will not try to update it even if versions.yaml change.

[amatas]

LGTM

[mrtyler]

LGTM :(

Although this is a temporary hack, I think it could confusion for our users in the short-term (cc @javihernandez @klown @cindyli).

Anyone else we should notify? Email to architecture@?

[natarajaya]

LGTM

[stepanstipl]

@mrtyler thanks, it will not change the behavior on already existing clusters - as the dataloader will not run anyway in a current state, so the only difference will be on newly created clusters.

[javihernandez]

Anyone else we should notify? Email to architecture@?

Hey @mrtyler, thanks for the heads up, I think it's fine as long as @cindyli and @klown (and probably @sgithens, since I know he's been using gpii-infra) are aware of it.
In any case, just to confirm, if I need to re-deploy the dataloader for whatever reason, I need to update this file with the correct sha in addition to the change in the versions.yml file, right? Also, given that this is a temporary solution, I assume that you'll let us know again when this is not required anymore?

[stepanstipl]

@javihernandez currently, if you would need to redeploy new data loader, it's correct you would have to update the sha at the same place this PR freezes it (gcp/modules/gpii-dataloader/main.tf) but also the dataloader as of now will only actually run once when creating a new cluster (or delete the dataloader batchJob from the existing cluster first).

Otherwise this is temporary only till we get to resolve GPII/universal#626 and related #163 PRs