Hi, what is the standard procedure to upgrade the ipfs-cluster service and the underlying ipfs service to use new versions.
For example, there’s now an issue with too many open files, there is a new go-ipfs version which solves this, how can I update the cluster peers without having downtime, sync or versioning issues?
Do I have to wait for the compliant ipfs-cluster version to be released or can I update only the peers directly? If possible to update only the underlying ipfs peers go-ipfs versions, how should this be done?
Hello,
The usual upgrade procedure is to stop all cluster peers, replace the binaries with the new ones, and start them again (as of latest release the daemon --upgrade flag is not used anymore). Currently, if it new version is only a patch release, you don’t have to do all the nodes at once. We will likely make this more flexible starting in the next release.
go-ipfs upgrades are independent. Unless go-ipfs breaks its API (something it usually does not do), it can be upgraded separately from cluster. go-ipfs can be restarted without touching anything in the cluster peers.
1 Like