add caching for GET /rounds/current #143
Conversation
| } | ||
| res.setHeader( | ||
| 'cache-control', | ||
| `max-age=${roundParam === 'current' ? 60 : 31536000}}` |
There was a problem hiding this comment.
I thought about a dynamic setting for the current round, which ensures it's not longer than the expected round end. Wdyt @bajtos?
There was a problem hiding this comment.
This is a good idea.
In #147, I reworked "get tasks for the current round" logic so that we don't need dynamic max-age any more.
However, we should eventually move the logic of determining "what's the current meridian round index" from spark-api to spark-checker. I can imagine that it can be useful for spark-checker nodes to not start any new work (retrieval) if the current round is coming to an end.
By the time we finish the retrieval, submit the measurement to spark-api, and spark-publisher commits the measurement to Meridian, the new round will have already started, and our measurement will be evaluated as invalid (measuring a task that does not belong to this round).
I think we should make the check robust to handle changes in round lengths. Ideally, the smart contract should indicate when the current round is expected to end. I think we already have this value in Meridian state - the public field currentRoundEndBlockNumber.
I am proposing to open a new GH issue to track this idea. WDYT?
|
Closing in favor of #147 |
Done. |
After merging, adjust https://dash.cloudflare.com/37573110e38849a343d93b727953188f/filspark.com/caching/cache-rules/7cc5c26ebc3c400c9666c5a0f83ac50d to read the
cache-controlheader again