panic: runtime error: slice bounds out of range [:2165545518] with capacity 158

[ldapster]

Question.

Hi, older version of ipfs cluster (gcpm 3 pods) and one of the the pods is throwing the above error. Below is the stack trace:

goroutine 581 [running]:
github.com/dgraph-io/badger.(*valueLog).Read(0xc00040abf0, {0x2c6, 0x9e, 0x69513d0}, 0x54?)
	github.com/dgraph-io/[email protected]/value.go:1249 +0x44f
github.com/dgraph-io/badger.(*Item).yieldItemValue(0xc0029a2c60)
	github.com/dgraph-io/[email protected]/iterator.go:169 +0x13e

From looking at the 1.6/value.go source it appears that the header vlen value might be corrupted.

var h header
h.Decode(buf)
n := uint32(headerBufSize) + h.klen
return buf[n : n+h.vlen], cb, nil

The other two IPFS pods seem to be running fine. We are in a quandary on how to fix this node?
Is it possible to maybe backup and restore somehow from one of the other pods?

[marius-enlock]

Hi,

I also randomly ran in my CI/CD tests on this runtime error, but I ignored it as it has very low probability and couldnt reproduce it