Default to retrying after errors forever.#125
Conversation
Codecov Report
@@ Coverage Diff @@
## master #125 +/- ##
==========================================
+ Coverage 67.66% 70.27% +2.61%
==========================================
Files 32 36 +4
Lines 1976 2493 +517
==========================================
+ Hits 1337 1752 +415
- Misses 570 646 +76
- Partials 69 95 +26
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
| # Number of retries to perform after a pipeline plugin error. | ||
| # Set to 0 to retry forever. | ||
| retry-count: 10 | ||
| retry-count: 0 |
There was a problem hiding this comment.
I'm personally against making this be the default. In a happy path setup for an experienced user this may very well make sense. But for a newbie starting out and with some configuration that's off the effect will be to have conduit hang instead of failing fast.
But I wasn't aware that this was the default for indexer 2.0 and others should express their thoughts.
There was a problem hiding this comment.
I was surprised too. The current default was made for pretty much the reason you're describing. Failing fast seems like a good thing. That said, several users brought up the behavior as something they appreciated about Indexer.
We don't retry during Init, so plugins have a chance to kill the app during startup if they want.
That said, I've considered adding some sort of retry for the importer to smooth over the case where Conduit starts before Algod.
Summary
On an errors, Indexer will retry forever. Users have expressed an appreciation for this behavior because they can perform maintenance on algod and it doesn't disrupt service on the downstream applications.
This changes Conduit's default retry on error behavior to have unlimited attempts, which matches how Indexer 2.x works.