Throttle IPFS_INIT_FAILED events

[lidel]

Problem

When user opts-in to sending usage metrics, and ipfs-webui is not able to connect to HTTP API, it will send a new IPFS_INIT_FAILED event to Countly every few seconds when it tries to reconnect and fails again. This produces multiple events sent every minute:

This behavior can be observed on https://webui.ipfs.io and inside of ipfs-desktop when a local node is shut down when webui is still open (need to opt-in to metrics first)

Additionally, on a slower machine ipfs-desktop shows ipfs-webui before go-ipfs is fully booted, but that triggers no more than one or two events.

Solution

We should be throttling how often IPFS_INIT_FAILED is sent.

Some quick ideas:

• We want to decrease noise of situations where node started within a few seconds (either automatically in ipfs-desktop or manually when user forgot to run it)
  • Idea: do not send send it immediately, but keep buffering events and wait for IPFS_INIT_FAILED state to remain for 10+ seconds.
• When IPFS_INIT_FAILED is confirmed by being real for over 10 seconds, we do not want to keep spamming events every few seconds.
  • Idea: send it only once every 5 minutes (memoize send function for 5min)

Tasks

Preview Give feedback

1. prevent sending more than 1 IPFS_INIT_FAILED event per user per app load
2. Ensure IPFS_INIT_FAILED events are being logged to error.log

[SgtPooki]

💯 We should separate the IPFS_INIT_FAILED metric from retry failures.

When I first saw the IPFS_INIT_FAILED metric, I thought it was supposed to only fire once for each user, for each app instance run, and I think that intuition should be codified. Retries continuing to fail are valid, but should be captured via a different metric.

[SgtPooki]

code search results: https://github.com/search?q=repo%3Aipfs%2Fipfs-webui%20IPFS_INIT&type=code

[ipfs-gui-bot]

🎉 This issue has been resolved in version 4.0.2 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀