Hermes Causes Osmosis Sentry to Crash When Accessing gRPC

[lightiv]

Crate

Hermes Relayer

Summary of Bug

Hermes Causes Osmosis Sentry to Crash When Accessing gRPC. gRPC is enable by default and the Osmosis Sentry node runs fine BUT when I enable Hermes to access the node it cause it to crash. I see that a transaction will output a long, long string of data and the node stops reading blocks.

Version

git version 2.25.1

Steps to Reproduce

For me, In stall the Osmosis sentry and then configure Hermes to access it. The problem occurs withing 5 minutes

Jul 21 05:35:24 osmosis-sentry-02 cosmovisor[2904]: 5:35AM INF Error reconnecting to peer. Trying again addr={"id":"00c328a33578466c711874ec5ee7ada75951f99a","ip":"35.82.201.64","port":26656} err="auth failure: secret conn failed: read tcp 94.16.108.152:40766->35.82.201.64:26656: i/o timeout" module=p2p tries=4
Jul 21 05:35:25 osmosis-sentry-02 cosmovisor[2904]: 5:35AM INF ABCIQuery data=0A087472616E7366657212096368616E6E656C2D301A0B18FFFFFFFFFFFFFFFFFF01 module=rpc path=/ibc.core.channel.v1.Query/PacketAcknowledgements result={"code":0,"codespace":"","height":"436445","index":"0","info":"","key":null,"log":"","proofOps":null,"value":"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...........

Acceptance Criteria

Hermes 6 is able to access gRPC for Osmosis witout killing the node

---

For Admin Use

• Not duplicate issue
• Appropriate labels applied
• Appropriate milestone (priority) applied
• Appropriate contributors tagged
• Contributor assigned/self-assigned

[adizere]

Hello,

Thanks for opening this issue, seems quite critical!

Could you please provide further info, concretely:

• how do you run Hermes? Is it hermes start, or some other command?
  • it may be useful to paste somewhere the whole Hermes log file so we can see what might have gone wrong
• can we also see you Hermes config.toml file?

[jackzampolin]

I've also experienced this. Requires restart of the node.

[lightiv]

I run Hermes using a service that starts it with hermes start

Do you have an e-mail address I can send the log and config.toml file to?

[andynog]

I've also experienced this. Requires restart of the node.

@jackzampolin are you also using cosmovisor to run the node ?

[ancazamfir]

gRPC should not crash full nodes so some fix needs to go into abci/SDK for that.
But we should fix hermes to not give the full node a cause to crash: Not seeing any logs attached here but the little snippet seems to indicate the crash happens on acknowledgment query. Maybe the output is too big. hermes uses pagination: ibc_proto::cosmos::base::query::pagination::all() in the request. I guess hermes could limit the number of items in the response by doing multiple requests/ pages.

[hu55a1n1]

Hi @lightiv, can you provide an update on this, please? Is this still a problem with the latest release?

[lightiv]

Hi Need to update the be latest binary which I will do today and get back with you. Thanks.

[lightiv]

I am running 7.0.1 and gRPC still kills the node which this output being the final gashp.

.

[lightiv]

Oh, wait. By the time I finished the above post the console had timed out and when I reconnected the node was running. I will let it go for a bit. The above is Osmosis. I will try Akash also.

[romac]

@lightiv Are you still seeing the same issue with Hermes 0.7.2? Even if we haven't gotten to using pagination yet, this version should be much lighter load on the nodes now, both at startup and during normal operations.

[lightiv]

I have had this running for a couple of hours on the Akash and Osmosis nodes having issues and they have not become unresponsive. The Akash node is running fine and appears to be keeping up with the current block height. The Osmosis node is crawling and continually slowing. It will not be able to get to the current block height.

[lightiv]

Oh, sorry need to refresh. Let me update to v0.7.2 and get back to you. I am currently on v7.0.1.

[lightiv]

I have upgraded to v0.7.3 and I am not having any issues with the Osmosis node crashing. The problem now is that hermes times out accessing gRPC:

Oct 18 03:23:03 hermes[661828]: Oct 18 03:23:03.701 ERROR ThreadId(01) skipped workers for connection connection-13 on chain osmosis-1, reason:
Oct 18 03:23:03 hermes[661828]:    0: relayer error
Oct 18 03:23:03 hermes[661828]:    1: error in underlying transport when making GRPC call
Oct 18 03:23:03 hermes[661828]:    2: transport error: error trying to connect: tcp connect error: Connection timed out (os error 110)
Oct 18 03:23:03 hermes[661828]:    3: error trying to connect: tcp connect error: Connection timed out (os error 110)
Oct 18 03:23:03 hermes[661828]:    4: tcp connect error: Connection timed out (os error 110)
Oct 18 03:23:03 hermes[661828]:    5: Connection timed out (os error 110)
Oct 18 03:23:03 hermes[661828]: Location:
Oct 18 03:23:03 hermes[661828]:    /home/runner/.cargo/registry/src/github.com-1ecc6299db9ec823/flex-error-0.4.3/src/tracer_impl/eyre.rs:31

I have verified that gRPC is accessible externally:

grpcurl -plaintext 0.0.0.0:9090 list

cosmos.auth.v1beta1.Query
cosmos.bank.v1beta1.Query
cosmos.base.reflection.v1beta1.ReflectionService
cosmos.base.tendermint.v1beta1.Service
cosmos.distribution.v1beta1.Query
cosmos.evidence.v1beta1.Query
cosmos.gov.v1beta1.Query
cosmos.params.v1beta1.Query
cosmos.slashing.v1beta1.Query
cosmos.staking.v1beta1.Query
cosmos.tx.v1beta1.Service
cosmos.upgrade.v1beta1.Query
grpc.reflection.v1alpha.ServerReflection
ibc.applications.transfer.v1.Query
ibc.core.channel.v1.Query
ibc.core.client.v1.Query
ibc.core.connection.v1.Query
osmosis.claim.v1beta1.Query
osmosis.epochs.v1beta1.Query
osmosis.gamm.v1beta1.Query
osmosis.incentives.Query
osmosis.lockup.Query
osmosis.mint.v1beta1.Query
osmosis.poolincentives.v1beta1.Query
testdata.Query

[adizere]

Can you please share with us the relevant section of your Hermes config.toml? It should look similar to this:

https://github.com/informalsystems/ibc-rs/blob/91a4d1d0db84fdcc7f780a775f8f25eb57412cf4/config.toml#L83

Also, what is the output of hermes health-check ?

[lightiv]

The issues I was having has been resolved by installing everything on one box. I am now relaying.