Yaroslav Dynnikov authored 2 years ago and Yaroslav Dynnikov committed 2 years ago

The error message about `on_shutdown` trigger failure recently changed
in tarantool [1]. Our test doesn't catch it anymore.

- [1] https://github.com/tarantool/tarantool/commit/ca59d3050c

When instance runs `postjoin` function, it's supposed to have both
target and current grade Offline. Taking a read barrier in that state
makes no sence. Instead, an instance should set target grade online
asap. Ensuring it syncs is the responsibility of the governor.

Excessive read barrier affects `restart_both` test. One of the instances
(i1 who becomes a follower) sends a `read_index` request to i2 (leader).
But i2 reveals  `self.commit_to_current_term()` condition isn't met and
drops the request. As a result, i1 is obliged to wait for 10s timeout
before retrying `read_index` request again, but the test timeouts even
sooner.

An important test case (5b5d25f3) went missing after migration from
luatest to pytest (b6630a41).

Conftest operates subprocesses in a different way than luatest helpers
did: luatest used to signal a signal process (supervisor), while
`instance.kill()` signals the whole process group.

Yaroslav Dynnikov authored 2 years ago and Yaroslav Dynnikov committed 2 years ago

Eliminate code duplication, use `define_str_enum` macro from tarantool-module.
Remove personalized `FromStr::Err` for every enum defined that way.

After adding grade incarnations restarting instances now causes them to
instantly become unfit for being a voter (this may change later). The
test in question however isn't really related to instances restarting
and disrupting the communication by breaking stored procedures is also
applicable.

Saving cache isn't instantaneous. Usually it takes about a minute.
If another job tries to restore cache at the same time, it fails
with the error "tar: Unexpected EOF in archive".

Now tar writes to a different filename with a random suffix, and then
renames it atomically.

CI on master branch always reuses the same archive. As a result, its
size gradually grows. It already takes more than 6GB, while it should be
about 3GB large. It makes restoring the cache slower and slower.

With this patch the cache is cleared every time in master branch.

Yaroslav Dynnikov authored 2 years ago and Yaroslav Dynnikov committed 2 years ago

Retain it in tests only

Georgy Moshkin authored 2 years ago and Yaroslav Dynnikov committed 2 years ago

Tarantool synchronous replication is broken in such a way, that local
& asynchronous transactions get rolled back upon synchronous
transaction rollback. This is critical for us, because we use local
spaces for the raft log, which should be orthogonal to synchronous
business spaces, but is not due to this behavior. Until it is fixed, we
will not use synchronous replication.

See https://github.com/tarantool/tarantool/issues/7903

Rewrite it from scratch to satisfy tests and to represent governor
logics in general.