refactor(apps/price_pusher): crash on RPC failures #1730
Conversation
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
1 Ignored Deployment
|
| version: objectResponse.data!.version, | ||
| }; | ||
| } else { | ||
| throw new Error("Failed to refresh object reference"); |
There was a problem hiding this comment.
I'm not sure how often this may happen, but if it fails frequently and causes crashes, there is a higher chance that we lock all the gas price objects
There was a problem hiding this comment.
During my experiments I think I've resolved the issue of locked gases. It seems that it happens when we reuse a gas coin upon crash/restarts. I changed our deployment configuration to never allow to live instances at the same time and wait 10s before starting a new instance to make sure older ones are settled. No locking issues since this change.
I'll also run this code for sui for a day or so before merging to see how it works.
guibescos
left a comment
guibescos
left a comment
There was a problem hiding this comment.
I debugged it a bit, found couple of bugs
ali-behjati
force-pushed
the
price-pusher/refactor
branch
from
87c3623 to
8de8d28
Compare
| logger: Logger | ||
| ): Promise<SuiObjectRef[]> { | ||
| const signerAddress = await signer.toSuiAddress(); | ||
| const signerAddress = signer.toSuiAddress(); |
There was a problem hiding this comment.
drive-by: linter
| pythPackageId, | ||
| pythStateId, | ||
| wormholePackageId, | ||
| wormholeStateId, | ||
| endpoint, | ||
| keypair, |
There was a problem hiding this comment.
drive-by: linter
| this.logger.info({ signatures }, "updatePriceFeed successful"); | ||
| } catch (err: any) { | ||
| this.logger.error(err, "updatePriceFeed failed"); | ||
| return; |
There was a problem hiding this comment.
this return doesn't seem to do anything
There was a problem hiding this comment.
lol it was there before and i removed it. added it back to minimize the changes.
This PR updates the price pusher code to cause a crash when the RPC has issues. The RPC issues can be either the target network or Hermes.
This also solves #1704.