You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Jan 13, 2021. It is now read-only.
This is a simply a public issue to track RTrade swapping out the underlying infrastructure of our service for TemporalX to give users of our publicly deployed Temporal service a better UX, and to make management of it better. It will also prevent common exploits being done against go-ipfs to drop performance.
This doesn't meant to use this repo you need TemporalX, and the integration with go-ipfs will still be available, but in an alternate branch.
Overview
Running garbage collection on our production IPFS nodes is beyond unfeasible to the point that nuking the data directory with rm -rf and resyncing via IPFS Cluster is faster than garbage collection. As indicated by the immediate closure of #7213 on go-ipfs and from a github search on go-ipfs, there doesn't appear to be any activity involved in making GC feasible at scale.
While it's been the plan to remove go-ipfs from our production deployment and use TemporalX it's clear this needs to be fast tracked, because:
rm -rf & resync is an annoying process, and increases disk IO, bandwidth, etc...
We can't run GC at all
TemporalX is much faster, and the UX of our service will be better
No need to constantly monitor the syncing IPFS node for problems
Overall easier management of our production infrastructure
Makes the service more competitive as it gives us a performance edge not possible by any other IPFS pinning service
To do this the following tasks need to be done:
Minimal ipfs node to expose IPFS HTTP API services
The minimal node will immediately send a pin request to TemporalX the moment an IPFS HTTP API request is done
This will allow us to expose IPFS HTTP API services as we do now without needing to run major IPFS nodes
This minimal node will also allow us to to easily continue with the "only hash" check we do now that prevents double charging users.
Cut a v3 release of Temporal which uses TemporalX instead of go-ipfs for
Modify ipfs-pump to enable pumping data out of go-ipfs into TemporalX
Note
This is a simply a public issue to track RTrade swapping out the underlying infrastructure of our service for TemporalX to give users of our publicly deployed Temporal service a better UX, and to make management of it better. It will also prevent common exploits being done against go-ipfs to drop performance.
This doesn't meant to use this repo you need TemporalX, and the integration with go-ipfs will still be available, but in an alternate branch.
Overview
Running garbage collection on our production IPFS nodes is beyond unfeasible to the point that nuking the data directory with
rm -rfand resyncing via IPFS Cluster is faster than garbage collection. As indicated by the immediate closure of #7213 on go-ipfs and from a github search on go-ipfs, there doesn't appear to be any activity involved in making GC feasible at scale.While it's been the plan to remove go-ipfs from our production deployment and use TemporalX it's clear this needs to be fast tracked, because:
rm -rf& resync is an annoying process, and increases disk IO, bandwidth, etc...To do this the following tasks need to be done: