irods / irods_docs

iRODS Documentation
3 stars 17 forks source link

Add words to compound resource section about checksums and archive resources #258

Open korydraughn opened 1 month ago

korydraughn commented 1 month ago

Replicas stored in an archive resource of a compound resource hierarchy cannot have checksums. This is due to the fact that some storage resources (e.g. tape, s3) do not support all POSIX operations.

Also, copying the checksum from the cache isn't sufficient because the replication can fail.

alanking commented 1 month ago

For ease of reference...

https://docs.irods.org/4.3.2/plugins/composable_resources/#compound

Current wording around this:

A replica created on an archive resource in a compound resource hierarchy by a sync-to-archive operation will not have its checksum calculated and no checksum will be applied to the replica's entry in the catalog. This is because the resource plugin may not support calculating the checksum or may be extremely expensive. Historically, the server has attempted to calculate the checksum on archive resources and an error is returned and caught interally: DIRECT_ARCHIVE_ACCESS. It was then waived away as a non-error and the replica on the archive resource received the checksum from the replica on the cache resource. This is no longer the case as the recorded checksum cannot be trusted. Therefore, no checksum is recorded for replicas synced to archive resources.

korydraughn commented 1 month ago

That text is exactly the target of this issue. I've re-read it a few times now and it does cover the topic.

The objective of this issue is now to consider improving the wording so there's no room for confusion. Adding one statement like the following is probably good enough.

trel commented 1 month ago

would, of course, need to be updated again after https://github.com/irods/irods/issues/3127 is completed.