Would possibly show up when a consumer leader changes for a consumer
that had redelivered messages and for instance messages were inbound
on the stream.
Resolves#2912
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
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>
Since the "next" timer value is set to the AckWait value, which
is the first element in the BackOff list if present, the check
would possibly happen at this interval, even when we were past
the first redelivery and the backoff interval had increased.
The end-user would still see the redelivery be done at the durations
indicated by the BackOff list, but internally, we would be checking
at the initial BackOff's ack wait.
I added a test that uses the store's interface to detect how many
times the checkPending() function is invoked. For this test it
should have been invoked twice, but without the fix it was invoked
15 times.
Also fixed an unrelated test that could possibly deadlock causing
tests to be aborted due to inactivity on Travis.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Removed the warnings, instead have a sync.Map where they are
registered/unregistered and can be inspected with an undocumented
monitor page.
Added the notion of "in progress" which is the number of messages
that have beend pop()'ed. When recycle() is invoked this count
goes down.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
We have seen situations where when a lot of pending messages accumulate,
there is a contention between the processing of the ACKs and the
checking of the pending map.
Decision is made to abort checking of pending list if processing of
ack(s) would be delayed because of that. The result is that a
redelivery may be post-poned.
Internally, the ACKs are also now using a queue to prevent processing
of them from the network handler, which could cause head-of-line
blocking, especially bad for routes.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Also added in more optimized version to select the first matching message in a message block for LoadNextMsg.
Signed-off-by: Derek Collison <derek@nats.io>
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>
If the consumer's sequence was not the same than the stream's sequence,
then the redelivery would always use the first duration from the
BackOff list.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
This allows a consumer to have exponential backoffs vs static AckWait and MaxDeliver.
When BackOff is set it will overridde AckWait to BackOff[0] and MaxDeliver will be len(BackOff)+1.
Signed-off-by: Derek Collison <derek@nats.io>
We will only send if all peers in our group are >= 2.7.1 and we will check for updates.
When a consumer follower takes over it will notify all pending requests that those requests are invalid now.
Signed-off-by: Derek Collison <derek@nats.io>
This should help with GC pressure, however, it may have an effect
on performance (based on some benchmark). Calling sync.Pool.Get/Put
too often has a performance impact...
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
Under load we could have a message committed to the underlying store when a consumer was being created and then it increase num pending again when the stream signals the consumers.
This fix just remembers the last seq of the state when we calculate sgap and test before adding in the stream code.
Signed-off-by: Derek Collison <derek@nats.io>
Cleaned up code, made more consistent, utilize loopAndGather.
Allow pull consumers to have AckAll as well as AckExplicit.
Signed-off-by: Derek Collison <derek@nats.io>
This allows stream placement to overflow to adjacent clusters.
We also do more balanced placement based on resources (store or mem). We can continue to expand this as well.
We also introduce an account requirement that stream configs contain a MaxBytes value.
We now track account limits and server limits more distinctly, and do not reserver server resources based on account limits themselves.
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>
The filestore would release a msgBlock lock while trying to load a cache block if it thought it needed to flush pending data.
With async false, this should be very rare but was possible after careful inspection.
I constructed an artificial test with sleeps throughout the filestore code to reproduce.
It involved having 2 Go routines that were through and waiting on the last msg block, and another one that was writing.
After the write, but before we flushed after releasing the lock we would also artificially sleep.
This would lead to the second read seeing the cache load was already in progress and return no error.
If the load was for a sequence before the current write sequence, and async was false, the cache fseq would be higher than what was requested.
This would cause the errPartialCache to be returned.
Once returned to the consumer level in loopAndGather, it would exit that Go routine and the consumer would cease to function.
This change removed the unlock of a msgBlock to perform and flush, ensuring that two cacheLoads would not yield the errPartialCache.
I also updated the consumer in the case this does happen in the future to not exit the loopAndGather Go routine.
Signed-off-by: Derek Collison <derek@nats.io>
If the interest existed prior to the initial creation of the
consumer, the gateway "watcher" would not be started, which means
that interest moving across the super-cluster after that would
not be detected.
The watcher runs every second and not sure if this is costly or
not, so we may want to go a different approach of having a separate
interest change channel that would be specific to gateways. But this
means adding a new sublist where the interest would be registered
and that sublist would need to be updated when processing GW RSub
and RUnsub?
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
When a consumer is configured with "meta-only" option, and the
stream was backed by a memory store, a memory corruption could
happen causing the application to receive corrupted headers.
Also replaced most of usage of `append(a[:0:0], a...)` to make
copies. This was based on this wiki:
https://github.com/go101/go101/wiki/How-to-efficiently-clone-a-slice%3F
But since Go 1.15, it is actually faster to call make+copy instead.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
