When we were optimizing for single cluster and large numbers of
leafnodes we inadvertently broke a daisy chained scenario where a server
was a spoke and a hub within a single hub server.
So interest on D would not propagate properly to server A as a
publisher.
```
B
/ \
A C -- D (SUB)
|
PUB
```
This records the server time when info for streams and consumers are
created so that tools such as the nats cli can calculate time deltas for
last ack, last delivered and so forth in the context of the server
clock.
This will help aleviate problems with client devices experiencing clock
jitter that can show up in user interfaces as negative seconds since
last ack etc
- [X] Tests added
- [X] Branch rebased on top of current main (`git pull --rebase origin
main`)
- [X] Changes squashed to a single commit (described
[here](http://gitready.com/advanced/2009/02/10/squashing-commits-with-rebase.html))
- [ ] Build is green in Travis CI
- [X] You have certified that the contribution is your original work and
that you license the work to the project under the [Apache 2
license](https://github.com/nats-io/nats-server/blob/main/LICENSE)
### Changes proposed in this pull request:
1: Improves error reporting for weighted mappings:
As it was, any error in a weighted mapping would return a very
unhelpfull error message.
e.g. `nats-server: mappingtest.cfg:38:39: interface conversion:
interface {} is []interface {}, not string`
This was because the line `err := &configErr{tk, fmt.Sprintf("Error
adding mapping for %q to %q : %v", subj, v.(string), err)}` would panic
on the `v.(string)` since in weighted mapping that interface{} is
actually a map[string]interface{} (since there's can be more than one
mapping in weighted mappings).
Now returns the actual error:
e.g. `nats-server: mappingtest.cfg:40:3: Error adding mapping for "bla"
: invalid mapping destination: wildcard index out of range in
{{wildcard(1)}}`
2: improves subject transform checking and catches if the destination is
using a mapping function and there are no partial wildcards in the
source.
By publishing retained messages on their own subject, we can apply max
messages per subject limits to them. Also include the domain name in the
subject to make consistent with the other MQTT streams.
Signed-off-by: Neil Twigg <neil@nats.io>
I was running a manual test moving from dev to this branch and
noticed that the consumer would receive only 1 message of the 10
messages sent as retained. So I modified the test to verify that
we receive them all and we did not.
The reason was that after the transfer we need to refresh the state
of the stream (stream info) since we attempt to load all messages
based on the state's sequences.
I have also modified a bit the code to update the MaxMsgsPer once
all messages have been transferred.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
While adding more complex test scenarios, some issues in
multiple-filters approach surfaced.
This PR adds those tests and addresses the issues.
Signed-off-by: Tomasz Pietrek <tomasz@nats.io>
As it was, any error in a weighted mapping would return a very unhelpfull error message.
e.g. `nats-server: mappingtest.cfg:38:39: interface conversion: interface {} is []interface {}, not string`
This was because the line `err := &configErr{tk, fmt.Sprintf("Error adding mapping for %q to %q : %v", subj, v.(string), err)}` would panic on the `v.(string)` since in weighted mapping that interface{} is actually a map[string]interface{} (since there's can be more than one mapping in weighted mappings).
Now returns the actual error:
e.g. `nats-server: mappingtest.cfg:40:3: Error adding mapping for "bla" : invalid mapping destination: wildcard index out of range in {{wildcard(1)}}`
2: improves subject transform checking and catches if the destination is using a mapping function and there are no partial wildcards in the source.
Signed-off-by: Jean-Noël Moyne <jnmoyne@gmail.com>
The issue really was that the test was sharing remote options. The way
options are used is not ideal since we reference the user provided
options (but it is relied upon now in many tests and possibly users
setups). The other side of the issue was that when no local account is
specified in a "remote" specification, we set it to the global account,
but that was done when creating the leafnode object (when soliciting),
which in the case of the test could race with the second server doing
the validation.
In this PR we move the setting to global account during the validation,
but also fixed the tests to not share the remote options configuration
slice between the two servers.
Resolves#4191
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
The issue really was that the test was sharing remote options. The
way options are used is not ideal since we reference the user provided
options (but it is relied upon now in many tests and possibly users
setups). The other side of the issue was that when no local account
is specified in a "remote" specification, we set it to the global
account, but that was done when creating the leafnode object (when
soliciting), which in the case of the test could race with the
second server doing the validation.
In this PR we move the setting to global account during the validation,
but also fixed the tests to not share the remote options configuration
slice between the two servers.
Resolves#4191
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
This is a rework of incorrect changes made in PR #4001. Changes did not
work for changes to export permissions. Test was modified to add export
changes.
This affects only the `dev` branch.
Signed-off-by: Ivan Kozlovic <ivan@synadia.com>
- [X] Branch rebased on top of current main (`git pull --rebase origin
main`)
- [X] Changes squashed to a single commit (described
[here](http://gitready.com/advanced/2009/02/10/squashing-commits-with-rebase.html))
- [ ] Build is green in Travis CI
- [X] You have certified that the contribution is your original work and
that you license the work to the project under the [Apache 2
license](https://github.com/nats-io/nats-server/blob/main/LICENSE)
### Changes proposed in this pull request:
Allows (exposes) the existing cluster filtering option of account level
subject mapping, thereby allowing you to define different mapping
destination per cluster (e.g. to insert the cluster name as a subject
token).
Example:
```
mappings {
"foo":[
{destination:"foo-east", cluster: "east", weight:100},
{destination:"foo-west", cluster: "west", weight:100},
]
}
```
This adds a workflow to mark issues and PRs stale after the configured
period of time, followed by closing the issue/PR after a subsequent
period of time if there was no additional activity.
The `debug-only` option is so currently, so even when merged, it will do
a dry-run and not perform any actions. Once we inspect the initial logs
of the effect of an initial run (impacting existing issues), we can
adjust accordingly and then follow-up with making it active.
For the debug logs to be enabled, we do need to add a repository secret
named `ACTIONS_STEP_DEBUG` with a value set to `true` per [this
instruction](https://github.com/marketplace/actions/close-stale-issues#debugging).
1. When catching up do not try forever and if needed reset cluster state.
2. In checking if a stream is healthy check for node drift.
3. When restarting a stream make sure the current node is stopped.
Signed-off-by: Derek Collison <derek@nats.io>
