Linearizability issue14890

[serathius]

cc @ahrtr @ptabor @geetasg @tjungblu @aphyr

Failed attempt to try reproducing #14890

I switched linearization model to append one resembling Jepsen https://jepsen.io/analyses/etcd-3.4.3#append

For now use single key and validates exact contents.
In loop each client:

• Read the value of key
• Create Txn if value didn't change, put original value appended with unique identifier.

This approach has one advantage that failed requests can be ignored. Model doesn't check exact requests, just ensures that prefix matches.

@aphyr From what I understood (don't know closure at all), one difference with append flow is that Jepsen compares modRevision in Txn instead value. Do you see any other differences or suggestions?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 21 days if no further activity occurs. Thank you for your contributions.

[serathius]

No longer needed with TXN tests present.