Test TestNoRaceJetStreamClusterCorruptWAL() would start to flap
because of the snapshot on cluster shutdown. Disable the snapshot
on exit for this test by stopping the raft node before shutdown.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- didRemove in applyMetaEntries() could be reset when processing
multiple entries
- change "no race" test names to include JetStream
- separate raft nodes leader stepdown and stop in server
shutdown process
- in InstallSnapshot, call wal.Compact() with lastIndex+1
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Use better indexing for lookups, we used to do simple linear scan backwards, now track first and last block.
Will expire the fss cache at will to reduce memory usage.
Signed-off-by: Derek Collison <derek@nats.io>
- Send snapshot only if leader
- When processing snapshot, start with a smaller inactivity interval
that will double up to 10sec or use 10sec directly once we get a
message. Reason for that is that it is possible that the request
for snapshot is sent while the leader has not yet setup the subscription
that receives the requests (or subscription has not fully reached the
cluster).
- Don't remember snapfile on err.
- Do not consider current if we have not had any activity.
- Stabilize stream scale up under active heavy publishing.
- Due to the publish pressure move the check for followers direct subs spinning up til after we stop publishing.
Signed-off-by: Derek Collison <derek@nats.io>
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
If there are more stream names that the current limit of 1024, getting
the list of names would return them all instead of using pagination.
For "stream infos", the Total amount returned would be the API limit
instead of the actual number of streams.
Resolves https://github.com/nats-io/natscli/issues/541
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
If a JS API request is received from a non client connection, it
was processed in its own go routine. To reduce the number of
such go routine, we were limiting the number of outstanding routines
to 4096. However, in some situations, it was possible to issue
many requests at the same time that would then cause those requests
to be dropped.
(an example was an MQTT benchmark tool that would create 5000
sessions, each with one QoS1 R1 consumer (with the use of consumer_replicas=1).
On abrupt exit of the tool, the consumers and their sessions needed
to be deleted. Since would cause fast incoming delete consumer requests
which would cause the original code to drop some of them)
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
There was an issue with MaxWaiting==1 that was causing a request
with expiration to actually not expire. This was because processWaiting
would not pick it up because wq.rp was actually equal to wq.wp
(that is, the read pointer was equal to write pointer for a slice
of capacity of 1).
The other issue was that when reaching the maximum of waiting pull
requests, a new request would evict an old one with a "408 Request Canceled".
There is no reason for that, instead the server will first try to
find some existing expired requests (since some of the expiration
is lazily done), but if none is expired, and the queue is full,
the server will return a "409 Exceeded MaxWaiting" to the new
request, and not a "408 Request Canceled" to an old one...
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
When deciding to compact a file, we need to remove from the raw
bytes the empty records, otherwise, for small messages, we would
end-up calling compact() too many times.
When removing a message from the stream, in FIFO cases we would
write the index every 2 seconds at most when doing it in place,
when when dealing with out of order deletes, we would do it for
every single delete, which can be costly. We are now writing
only every 500ms for non FIFO cases.
Also fixed some unrelated code:
- Decision to install a snapshot was based on incorrect logical
expression
- In checkPending(), protect against the timer being nil which
could happen when consumer is stopped or leadership change.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- Remove code coverage from Travis and add it to a GitHub Action
that will be run as a nightly.
- Use tag builds to exclude some tests, such as the "norace" or
JS tests. Since "go test" does not support "negative" regexs, there
is no other way.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
The main issue was that in mixed-mode, the interest through gateway
may still be in optimistic mode, which when creating the source
consumer would start delivery before we had a chance to setup
the subscription to receive those messages.
The approach is to create the subscription prior to sending
the consumer create request. Also refactored a bit the code in
the hope to make the retries a bit more bullet proof.
We may also look at making sure that gateways are switched to
interest-mode when detecting a mixed-mode setup.
Also fixed a defect that could cause a source to be canceled
when updating a stream.
Resovles #2801
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- Possibly missing some early messages from the sourced stream
- In some cancel situations, the processing of sourced messages
would not longer work
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- Wait of some sort of routing to be in place before starting
the raft run loop
- Remove use of lock in apiDispatch that was not necessary but
could have cause a route to block, causing memory growth, etc..
Unrelated rename of some tests so that they start with TestJetStream
and TestJetStreamCluster for cluster tests, fixed some flappers
and ensure that tests that change RAFT timeouts put them back
to default values on exit.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- A stream could become leader when it should not, causing
messages to be lost.
- A catchup could stall because the server sending data
could bail out of the runCatchup routine but still send
the EOF signal.
- Deadlock with monitoring of Jsz
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Signed-off-by: Derek Collison <derek@nats.io>
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Also fixed a bug that could cause memory based replicated consumers to no longer work after snapshots and server restarts.
The snapshot logic would allow non-state changing updates to continously grow the raft logs. We also were too conservative on when we snapshotted and why.
Also added in ability to have FileStore.Compact() reclaim space from the block file from the head of last changed block.
Signed-off-by: Derek Collison <derek@nats.io>
During contention to the head write blk, the system could perform worse memory wise compared to simple go runtime.
Also had some references for the subject of messages bloating memory.
Signed-off-by: Derek Collison <derek@nats.io>
Previously we would rely more heavily on Go's garbage collector since when we loaded a block for an underlying stream we would pass references upward to avoimd copies.
Now we always copy when passing back to the upper layers which allows us to not only expire our cache blocks but pool and reuse them.
The upper layers also had changes made to allow the pooling layer at that level to interoperate with the storage layer optionally.
Also fixed some flappers and a bug where de-dupe might not be reformed correctly.
Signed-off-by: Derek Collison <derek@nats.io>
For reason explained in previous commit, for tests that were
expecting the number of ack/pending to be of a certain value after
an Ack(), they would be flapping. Replaced all references and
we can go back to selectively call Ack() when AckSync() is not
needed.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Also had to change all references from `path.` to `filepath.` when
dealing with files, so that it works properly on Windows.
Fixed also lots of tests to defer the shutdown of the server
after the removal of the storage, and fixed some config files
directories to use the single quote `'` to surround the file path,
again to work on Windows.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
When a stream has multiple subjects and a consumer filters the stream to a small and spread out list of messages the logic would do a linear scan looking for the next message for the filtered consumer.
This CL allows the store layer to utilize the per subject info to improve the times.
Signed-off-by: Derek Collison <derek@nats.io>
This has nothing to do with the test, we just want to make sure the leader steps down and there were no low level errors on the fetch.
Signed-off-by: Derek Collison <derek@nats.io>
Currently this code returns a 200 and { "status": "ok" } iff all configured ports are open
and if JetStream is configured and we have contact with the metaleader and the cluster and all streams are up to date.
Signed-off-by: Derek Collison <derek@nats.io>
Under certain situations large number of consumers that are racing to update state or delete their stores during a delete
would start taking up OS threads due to blocking disk IO. When this happened and their were a bunch of Go routines becoming
runnable the Go runtime would create extra OS threads to fill in the runnable pool and would exhaust the max thread setting.
This code places a channel as a simple semaphore to limit the number of disk IO blocking OS threads.
Signed-off-by: Derek Collison <derek@nats.io>
Actually faster to not track at all and generate on the fly. Saves lots of memory too.
When we update the stream state to include runs, etc will update this as well.
Signed-off-by: Derek Collison <derek@nats.io>
