Cherry-pick PRs for v2.10.14-RC.2 #5278
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
neilalexander
force-pushed
the
neil/release/v2.10.14
branch
from
c204c49 to
f7f7a09
Compare
Signed-off-by: depthlending <[email protected]>
…dexCacheBuf. Signed-off-by: Derek Collison <[email protected]>
The code was just checking if max_deliver is >= than backoff, however infinite max_deliver is represented by -1, so it was erroring. This fixes that behavior. Signed-off-by: Tomasz Pietrek <[email protected]>
Signed-off-by: Waldemar Quevedo <[email protected]>
Signed-off-by: Waldemar Quevedo <[email protected]>
The main issue for router/gateway connections were that they would be registered with the global account but never removed from there, which would cause the connection to be retained. The memory leak was apparent in one user setup with large amount of subscriptions that were transfered to a slow gateway, causing accumulation of partial buffers before the connection was dropped. For gateway connections, we also needed to clean the outbound's map since it can contain sublist referencing its own connection. Same for subscriptions in c.subs. Another found cause for retained connection is with leafnode where we are using a subscription referencing the leafnode connection that was globally registered in a global sublist but never removed. Also the connection's c.out.sg (a `*Cond`) needed to be set to `nil` when the connection is closed. Without that, it seems that the connection would not be released (at least based on experimentation). We now make sure that c.out.sg is not nil before sending a signal. The bottom line is that it looks like having an object referencing itself in some way, prevents the GC from releasing the object, even if the "top level" object is no longer reachable. For instance, suppose `obj.someField = obj` seem to prevent the GC from releasing `obj`, even if `obj` itself is no longer referenced. The original issue/PR (#5212) was simply cleaning the c.out.wnb, but the way it was done was racy since c.out.wnb is used without lock in flushOutbound. Once the retain/release issue is fixed, cleaning this buffer is not really required (but good practice especially if not running with GOMEMLIMIT), so we take care of cleaning this up, but under the protection of the flushOutbound flag. If set, flushAndClose() will not do the cleaning, flushOutbound will do it. Relates to #5212 Signed-off-by: Ivan Kozlovic <[email protected]>
Failed reads here might result in duplicate nonces. Signed-off-by: Neil Twigg <[email protected]>
Otherwise we might not overwrite the message with random bytes properly. Signed-off-by: Neil Twigg <[email protected]>
In this case, an empty-but-non-nil map would not be copied, which could result in the clone referring to the original map. Also implement a deep copy of `AllowedConnectionTypes` for `NkeyUser`, like `User` already has. Signed-off-by: Neil Twigg <[email protected]>
…r stream and has a max messages limit and has discard policy new. The stream should never go over the limit but should still get all the messages in the stream being sourced (i.e. getting more messages as they are consumed from the WQ stream) Signed-off-by: Jean-Noël Moyne <[email protected]>
Signed-off-by: Neil Twigg <[email protected]>
Tests were executed against the parent commit of when they were added to make sure that the tuning is not affecting the reproducibility of the test. Note that some of the tests were sometimes not detecting the bug, even when running with `-count=xxx`, so in that case, it is difficult to know if the tuning would prevent detection if the bug were to be reintroduced. The FileStore and MsgTrace tests were also added as a matrix run in Travis/GA. Signed-off-by: Ivan Kozlovic <[email protected]>
Signed-off-by: Ivan Kozlovic <[email protected]>
Signed-off-by: Ivan Kozlovic <[email protected]>
Excluded the files that were auto-generated (with a do not edit comment). Signed-off-by: Ivan Kozlovic <[email protected]>
…store if upto sequence was not found, e.g. deleted already. Signed-off-by: Derek Collison <[email protected]>
We were getting partial or no cache errors when hitting it from multiple go routines in quick succession. Signed-off-by: Derek Collison <[email protected]>
…valid. If we think we need to calculate a new first sequence but the one we have is still valid, this could lead to messages being skipped and not sent to a consumer. Signed-off-by: Derek Collison <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
…raddles multiple allocations in go 1.22.1 on darwin. Signed-off-by: Derek Collison <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
Specifically around interest and limited streams that are clustered. Since we operate consumers and streams independently, when a stream is removing msgs when acked and acks are coming into different servers at different times, etc. Messages can get stranded. We introduce a better check of the interest state based on a collective low ack floor that will purge our copy of the stream to catch up. We run this periodically in monitorStream and during certain conditions. Thanks to Wally for introducing the test that stresses this part of the system. Signed-off-by: Derek Collison <[email protected]>
For the example hardened systemd file, which is derived from config files which Synadia uses in production, add more examples: * Show setting GOMEMLIMIT env var + but also nudge towards using the external environment file for it * Use an external environment config file if it exists * Show mount-point dependency for JetStream * Configure automatic restarts * Show using InaccessiblePaths to lock down FS access even further * Setup a service alias of simply 'nats'.
Signed-off-by: Neil Twigg <[email protected]>
Signed-off-by: Derek Collison <[email protected]>
…ects. Signed-off-by: Derek Collison <[email protected]>
It concerns me a little that by duplicating the same test for both file and memory stores, that differences could slip in later that mean we are no longer testing for consistency between the two stores. This approach allows us to write tests that only use the `StreamStore` API and have them tested against all store permutations. Signed-off-by: Neil Twigg <[email protected]>
Signed-off-by: Neil Twigg <[email protected]>
neilalexander
force-pushed
the
neil/release/v2.10.14
branch
from
f7f7a09 to
e3c9236
Compare
… unprotected. Signed-off-by: Derek Collison <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Cherry-pick the following PRs into the v2.10.14 release branch:
rand.Read,eraseMsgin filestore #5248clonebehaviour ofUserandNkeyUser#5246LoadNextMsgMulti#5275Signed-off-by: Neil Twigg [email protected]