Open henrikjohansen opened 6 years ago
We're working on CRC32C for all data on-disk. It should arrive soon.
The segment files now have multiple CRC32C checksums inside. For this feature we need to store a checksum of those in global for each segment. And perhaps an option to enable doing a stronger checksum (e.g. sha512) on the completed segment file and store that in global and maybe also publish the (id, sha) pairs to some stream external to Humio.
In order to detect manipulation of log data on-disk Humio should really have the ability to checksum the blocks that are written on-disk and to compare this against a known good checksum for those data block.
This would also yield some capabilities for detection storage error conditions such as bit rot, phantom writes / reads, etc which is important for long-term storage of log data.