Ferenc Szili

@mykaul The PR #19789 and the design for making TRUNCATE TABLE safe with tablets was rejected. I am currently implementing TRUNCATE as a topology operation. It will not be ready...

I made a draft design document listing the currently proposed solutions to TRUNCATE TABLE: https://docs.google.com/document/d/1o9JO2a9eS7auLFMEngO74d_k6Gg0B1iiH593GaOKU3Y

> #### 🔴 CI State: FAILURE > ✅ - [Build](https://jenkins.scylladb.com/job/scylla-master/job/scylla-ci/10530/) ✅ - [Unit Tests Custom](https://jenkins.scylladb.com/job/scylla-master/job/scylla-ci/10530/) **The following new/updated tests ran 100 times for each mode:** 🔹 topology_custom/test_truncate_with_tablets ✅ - [Container...

> @nodep does this still suffer from livelock while migration is going on? Yes, I am afraid it does. But I don't think this is a livelock, it is more...

> @nodep - did we test it with SCT longevity run? No, I have not done SCT longevity runs yet.

New in v2: Commits from previous version have been split: adding `storage_service` dependency to `storage_proxy` and the new implementation of `storage_proxy::truncate_blocking()` are now separate commits. Also, coroutinizing `truncate_statement::execute()` and adding...

> @nodep It's customary to summarize what changed in the new version so that reviewers know what to focus on. For example like this: [#19084 (comment)](https://github.com/scylladb/scylladb/pull/19084#issuecomment-2147169674) I've added a summary...

New in v3: - Changed `storage_service::get_erm_when_table_has_no_migrations` to allow vnode tables. This way we avoid situation with truncate table concurrently with bootstrap. Truncate table will now follow the same code path...

> Is the code vulnerable to livelocks? If some tablets are being continuously migrated, what is the latency of TRUNCATE? Yes, it is vulnerable to long latencies on tables with...

This seems to be one of those dreaded "It happens on my machine". Do you have any ideas what I could check? For me, it fails every time, unless I...