nats-io / nats-streaming-server / 2066

92%

main: 92%

Pull #923

travis-ci

[FIXED] Clustering: Snapshot skip sequence on error

If the leader had an issue sending a requested message, it would
simply log the error but continue with the next. The consequence
would be that the follower would get a "gap" and empty the store
up to that sequence assuming this was the new first available
message.
The missing `return` statement has been added, but there would
still be a situation where the follower could wrongly assume that
the mismatch in sequence between the request and the response
meant that this was the new first available message in the leader.
The leader is now sending - through the reply subject - information
that indicates that this is the first message. If not, the follower
has to verify that when getting a mismatch sequence, if this is
because messages were lost or because this is indeed the first
available. It is possible through a request of a single message
as opposed to a batch.

Also replaced some panic() calls due to Unmarshal() where a returned
error is more appropriate.

Resolves #921

Signed-off-by: Ivan Kozlovic <ivan@synadia.com>

51 of 51 new or added lines in 2 files covered. (100.0%)

11930 of 12971 relevant lines covered (91.97%)

50283.98 hits per line