Retrievals not working

[dkkapur]

Likely due to lotus version upgrade causing changes to the retrieval RPC call, submitted retrievals via the endpoint are instantly failing with the following error:

"ErrorMessage": "RPC error (-32601): method 'Filecoin.ClientRetrieveWithEvents' not found",

We should fix this or revert back to an older lotus version per discussion in sync on Dec 3.

[willscott]

It looks like deployments currently re-deploy lotus with whatever's at https://hub.docker.com/r/coryschwartz/lotus/tags

Would be great if we can pin it to lotus v1.13.1 for now

[coryschwartz]

The image tag is set through the controller API when the bot is created

https://github.com/filecoin-project/dealbot/blob/main/controller/spawn/spawner.go#L24

By default, it's the filecoin repo, not the coryshwartz repo

with that said, I can edit the image to move it to whichever version we need.

[willscott]

it looks like that is linked against https://github.com/filecoin-project/helm-charts/blob/master/charts/lotus-bundle/values.yaml#L107

is there an easy way to pin lotus, rather than ending up with whatever is in 'nightly'?

[willscott]

I see a place to specify lotus tag in dealbot, actually.

@Deep - we should try making a bot where that bottom field on the controller is not the default 'nightly', but rather 'v1.13.1'

On Tue, Dec 7, 2021 at 1:30 PM Will Scott @.***> wrote:

it looks like that is linked against https://github.com/filecoin-project/helm-charts/blob/master/charts/lotus-bundle/values.yaml#L107

is there an easy way to pin lotus, rather than ending up with whatever is in 'nightly'?

[dkkapur]

Just tried to deploy one, lets see if it works! Left everything else the same as the other slingshot bots I deployed earlier, just added the tag v1.13.1.

[dkkapur]

@willscott - does this fix the RPC call issue or still requires additional wiring for retrievals to work? I have not been tracking the changes for v1.13

[willscott]

It should not cause "method 'Filecoin.ClientRetrieveWithEvents' not found" errors 🤞

[coryschwartz]

oh wow, @willscott I think that's actually a bug. We are deploying nightly regardless of what we supply throught he API.

@dkkapur I made sure you're going to be running lotus v1.13.1 on that node, since that is clearly the intention. but it did deploy with lotus:nightly just now.

[dkkapur]

@coryschwartz thanks - can confirm that some of the retrievals I manually queued up today morning worked.

Waiting for feedback from Slingshot side to see if bulk submissions via endpoint are going through fine as well.

[dkkapur]

Rollback worked. Slingshot deal submissions worked. Closing this out for now, tracking remaining items in linked issues.