When the option Cluster.NoAdvertise is false, a server will send
an INFO protocol message to its client when a server has joined
the cluster.
Previously, the protocol would be sent only if the
joining server's "client URLs" (the addresses where clients connect
to) were new. It will now be sent regardless if the server joins
(for the first time) or rejoins the cluster.
Clients are still by default invoking the DiscoveredServersCB callback
only if they themselves detect that new URLs were added. A separate
PR may be filled to client libraries repo to be able to invoke
the callback anytime an async INFO protocol is received.
Based on @madgrenadier PR #597.
If server requires TLS and clients are connecting, and a Connz
request is made while clients are still in TLS Handshake, the
call to tls.Conn.ConnectionState() would block for the duration
of the handshake. This would cause the overall http request to
take too long.
We will now not try to gather TLSVersion and TLSCipher from a
client that is still in TLS handshake.
Resolves#600
The http servers for those two were recently modified to set
a ReadTimeout and WriteTimeout. The WriteTimeout specifically
caused issues for Profiling since it is common to ask sampling
of several seconds. Pprof code would reject the request if it
detected that http server's WriteTimeout was more than sampling
in request.
For monitoring, any situation that would cause the monitoring code
to take more than 2 seconds to gather information (could be due
to locking, amount of objects to return, time required for sorting,
etc..) would also cause cURL to return empty response or WebBrowser
to fail to display the page.
Resolves#600
Using a goto based loop makes it become a leaf function which can be
inlined, making us get a slight performance increase in the fast path.
See: https://github.com/golang/go/issues/14768
This is just for tests since from main.go, the FlagSet is set
with ExitOnError so Parse() would call os.Exit(2).
Regardless, I wanted to add error checking and a test for that.
Related to #578
The removal of SetClientAuthMethod removed any possibility of providing
a custom auth backend.
This patch add it back as a Option attribute, so we can wait comfortably for #434,
which aims to provide more extensible external Auth.
There were some cases where override would not work. Any command
line parameter that would be set to the type default value (false
for boolean, "" for string, etc) would not be taken into account.
I moved all the flags parsing and options configuration into
a new function, which may help reduce code duplication in
NATS Streaming.
The other advantage of moving this in a function is that it
can now be unit tested.
I am also removing call to `RemoveSelfReference()` which attempted
to remove a route to self, which has been already solved at runtime
with detecting and ignoring a route to self.
This function would be invoked only when routes were defined in
the configuration file, not in the command line parameter.
Removing this call also solves an user issue (#577)
Resolves#574Resolves#577
This is similar to #561 where `*` and `>` characters appear in tokens
as literals, not wilcards.
Both Insert() and Remove() were checking that the first character
was `*` or `>` and consider it a wildcard node. This is wrong. Any
token that is more than 1 character long must be treated as a literal.
Only for token of size one should we check if the character is `*`
or `>`.
Added a test case for Insert and Remove with subject like `foo.*-`
or `foo.>-`.
The issue was that a subject such as `foo.bar,*,>` would be
inserted to the cache as is, but when trying to remove from the
cache, calling matchLiteral() with the above subject in the cache
against the same subject would return false. This is because
matchLiteral would treat those characters as wildcards token.
Note that the sublist itself splits subjects on the `.` separator
and seem not bothered by such subject (would have `foo` and `bar,*,>`
tokens). Also, note that IsValidSubject() and IsValidLiteralSubject()
properly checked that the characters `*` and `>` are treated
as wildcards only if they are tokens on their own.
Resolves#558
- Move the kill of a server in a cluster test to ensure that
list of routes to remove is not empty.
- Change write_deadline reload value to 3s to make it different
from default value
- Add test for option that does not support hot-swapping
- Created a setter for the closed flag.
- Check if route is closed under lock and set a boolean if so,
so we don't check c.route outside of c's mutex.
- Ensure that we do not create a route on shutdown, which would
leave a connection hanging (was seen in some config reload tests).
The use of the `svc` API prevented the NATS Server to run as
a container on both nanoserver and windowsservercore Docker images.
An attempt was made to replace svc.Debug with normal server.Start()
if detected to be interactive, however, that did not work. The
fact of detecting if interactive or not already requires connecting
to the service controller apparently.
This change looks up for an environment variable (NATS_DOCKERIZED)
and if set to "1", will not make use of the `svc` package.
This environment variable will be set in the Docker image (in
nats-docker/windows/nanoserver/Dockerfile and windowsservercore/Dockerfile).
Resolves#543
When A connects to B and B connects to A (either based on static
configuration - explicit routes, or because of auto-discovery -
implicit routes), it is possible that each server initially
registers the route from the opposite TCP connection. It will
then result in each server dropping the connection.
We were previously setting a retry flag in the first accepted route
based on the name of servers, which means that regardless of
duplicate detection, the server with the "smaller" server name would
try to reconnect when the route connection was closed. For instance,
suppose that server B connects to server A, when B disconnects, A
would try to reconnect once to B. This became problematic in the
case of configuration reload, because removing the route from B to
A would still result in a route created from A to B.
Also, when a route attempts a reconnect, a random delay is added
to avoid repeated failure cycles that may occur in case where
A connects to B and B to A.
It's hard to implement a bulletproof solution for cleaning up the
symlinks created by config reload tests on failure since, for example,
there is nothing we can do when log.Fatal is called. Instead, prevent
the existence of a symlink from failing the tests if there is one
hanging around. Generally, these symlinks will not be left unless
os.Exit was called (as is the case with log.Fatal).
