Restart IPFS daemon once every 24 hours as it is leaking memory

[burdakovd]

ipfs/kubo#3318

[whyrusleeping]

The daemon shouldn't be leaking memory, what version are you using?

Could you grab a heap dump profile of the daemons memory usage for us? See this doc for a how-to: https://github.com/ipfs/go-ipfs/blob/master/docs/debug-guide.md

[burdakovd]

As I mentioned here, the version is docker image jbenet/go-ipfs:latest, 0.4.17.

Here is some profiling data from daemon that has been running for 8 days (in docker on a machine with 1.7 GB RAM and 1 GB swap, not the same machine where I initially observed high consumption).

Currently it appears to be using 1661 MB virtual memory and 712 MB resident memory.

Hope this helps: https://cloudflare-ipfs.com/ipfs/QmckqEGDeYP6qtmBg1HgFd5VWGWJXNuusKoEB4Pz4kwaFg/
Or same in archive: https://cloudflare-ipfs.com/ipfs/QmPSK6DT4FhsQbx465d4eDpcU6s5x8JVH47k17xtTo2Nbw/

[Stebalien]

This looks like libp2p/go-libp2p#419 (recently fixed, release TBD). However, that doesn't account for everything. The heap really isn't that large. I wonder if we're getting memory fragmentation in the heap.

I'll go ahead and cut a release as we have a bunch of goodies we should get out the door.