Durable changes made via the management API (e.g. adding
security-settings, adding address-settings, adding diverts) can be
reverted when reloading the XML at runtime.
This is a follow-up from ARTEMIS-2322.
The changes related to expired message are only there because
QueueFilterPredicate had a bug where the rate was correlated to expired
messages. When I fixed that I noticed that expired messages was actually
missing so I added it.
Casting the result of getPeerCertificates() to X509Certificate[] mirrors
what is done in the ActiveMQ "Classic" code-base.
A few tests which were imported from ActiveMQ "Classic" to verify our
OpenWire implementation were removed as they relied on a "stub"
implementation of javax.net.ssl.SSLSession that never would have worked
across multiple JDKs once javax.security.cert.X509Certificate[] was
removed. Furthermore, the tests appeared to be related to the OpenWire
*client* and not relevant to our broker-side implementation.
Aside from adding audit logging for message acknowledgement this commit
also consolidates the two nearly identical acknowledge method
implementations in o.a.a.a.c.s.i.QueueImpl. This avoids duplicating
code for audit logging, plugin invocation, etc. There is no semantic
change.
Due to the multi-threaded AMQP implementation the ThreadLocal variables
used by the AuditLogger to track the username and remote address don't
work properly. Changes include:
- Passing the audit Subject (set during authentication) and the remote
address explicitly for audit logging on the relevant ServerSession
methods rather than relying on the AuditLogger's ThreadLocal
variables
- Audit logging core session creation *after* successful authentication
so that we have the proper Subject; this is especially important for
the SSL certificate authentication use-case
- Renaming some methods and variables in AuditLogger to more accurately
reflect their intended use
- Adding JavaDoc and refactoring the getCaller methods on AuditLogger
- Refactor audit log testing and add a new test
As a follow-up to #3618/dc7de893747b90b627d729f9f18a758bb4dad9d5 update
checkstyle to the latest version, restoring the originally intended
"RightCurly" style, and updating all the code to properly adhere to the
style as enforced by the new checkstyle version.
The version of checkstyle we used before the aforementioned commit had
a bug which didn't properly enforced our intended "RightCurly" style
(see https://github.com/checkstyle/checkstyle/issues/6345). That commit
changed the style to accommodate the handful of unintended style
violations. This commit reverts that change for 2 main reasons:
- The style was always intended to use `alone` for both `METHOD_DEF`
and `CTOR_DEF`.
- There are over 1,000 existing uses of the intended style and around
30 violations of this style which were unintentionally allowed.
Reverting the style back to the original and cleaning up the unintented
violations makes the code more consistent and prevents further style
inconsistencies in the future.
There were a handful of other changes related to checkstyle bugs which
allowed unintended style violations. These were related to indentation
levels.
This closes#3619
(with some minor changes from Robbie to fix remaining violations)
The filter and the view use different convection for field names, ie the
connection view uses the `sessionID` field name while the connection filter
uses the `SESSION_ID` field name. This commit replace the field names used
by the filter with the field names used by the view preserving the backward
compatibility.
The provider of an SSL key/trust store is different from that store's
type. However, the broker currently doesn't differentiate these and uses
the provider for both. Changing this *may* potentially break existing
users who are setting the provider, but I don't see any way to avoid
that. This is a bug that needs to be fixed in order to support use-cases
like PKCS#11.
Change summary:
- Added documentation.
- Consolidated several 2-way SSL tests classes into a single
parameterized test class. All these classes were essentially the same
except for a few key test parameters. Consolidating them avoided
having to update the same code in multiple places.
- Expanded tests to include different providers & types.
- Regenerated all SSL artifacts to allow tests to pass with new
constraints.
- Improved logging for when SSL handler initialization fails.
Previously, when a session was reattached, all the close/failure listeners
were removed from the old connection and set onto the new connection.
This only worked when at most 1 session of the old connection was
transferred: When the second session was transferred, the old
connection already didn't contain any close/failure listeners anymore,
and therefore the list of close/failure listeners was overwritten by
an empty list for the new connection.
Now, when a session is being transferred, it only transfers the
close/failure listeners that belong to it, which are the session itself
+ the TempQueueCleanerUppers.
Modified a test to check whether the sessions are failure listeners of
the new connection after reattachment.
During session reattachment, also set the new connection on the
"session" member of the ServerSessionPacketHandler. Until now,
the connected ServerSessionImpl instance still referenced the old
connection although it had already been transferred on the new connection.
Change summary:
- Remove the existing Xalan-based XPath evaluator since Xalan appears
to be no longer maintained.
- Implement a JAXP XPath evaluator (from the ActiveMQ 5.x code-base).
- Pull in the changes from https://issues.apache.org/jira/browse/AMQ-5333
to enable configurable XML parser features.
- Add a method to the base Message interface to make it easier to get
the message body as a string. This relieves the filter from having
to deal with message implementation details.
- Update the Qpid JMS client to get the jms.validateSelector parameter.
WARNING: the eclipse static analyser is pretty limited and it cannot cope with cases like: correlated variables, exception paths, .... 70% of eclipse warnings on artemis codebase is a false alarm.
Anyway some variable correlations can be eliminated and code becomes more readable for humans too.
For "never-throws", the assumption is made explicit. If you disagree with the reverse-engineered assumption then it is likely an indication of a true potential NPE.
Last but not least, copy&paste is a common source of bugs. I suspect eclipse indirectly detected one such case.
Hope it helps
The fallback consumer authorization implemented in ARTEMIS-592 needs to
check for an *exact* security-settings match otherwise in certain
configurations a more general and more permissive setting might
be used instead of the intended more specific and more restrictive
setting.
The merge method in AddressSettings should *not* use any getters. It
should reference the relevant variables directly. Using any getters will
return default values in the underlying value is null. This can cause
problems for hierarchical settings.
Also fixed a few potential NPEs exposed by the test-case.
The existing deactivation callback happens *after* several important
services are shutdown (e.g. the remoting service which allows client
connectivity). This commit adds a new callback which is invoked *before*
any services are stopped. This is useful for embedded use-cases where
applications want to stop gracefully before any part of the broker is
stopped.
A default, empty method implementation is provided so that existing
callback implementations don't need to change.
This is to avoid shutting down the server on a critical failure in case the message is a few bytes shy
from beyond the max buffer size.
This will prevent the issue.
When deleting a durable scheduled message via the management API the
message would be removed from memory but it wouldn't be removed from
storage so when the broker restarted the message would reappear.
This commit fixes that by acking the message during the delete
operation.
Using a ThreadLocal for the audit user information works in most cases,
but it can fail when dispatching messages to consumers because threads
are taken out of a pool to do the dispatching and those threads may not
be associated with the proper credentials. This commit fixes that
problem with the following changes:
- Passes the Subject explicitly when logging audit info during dispatch
- Relocates security audit logging from the SecurityManager
implementation(s) to the SecurityStore implementation
- Associates the Subject with the connection properly with the new
security caching
- Adding a paragraph about addressing and distinct queue names
- Renaming match on peers, senders and receivers as "address-match"
- Changing qpid dispatch test to use a single listener
- Fixing reconnect attemps message
This commit is fixing:
- a missing commit that can make leak a connection
- restricting default specific commons-dbcp2 to the default data source
- setting poolPreparedStatements true by default
- configured embedded Derby to be in-memory to speedup tests
It add additional required fixes:
- Fixed uncommitted deleted tx records
- Fixed JDBC authorization on test
- Using property-based version for commons-dbcp2
- stopping thread pool after activation to allow JDBC lease locks to release the lock
- centralize JDBC network timeout configuration and save repeating it
- adding dbcp2 as the default pooled DataSource to be used
Replaces direct jdbc connections with dbcp2 datasource. Adds
configuration options to use alternative datasources and to alter the
parameters. While adding slight overhead, this vastly improves the
management and pooling capabilities with db connections.
This reverts commit dbb3a90fe6.
The org.apache.activemq.artemis.core.server.Queue#getRate method is for
slow-consumer detection and is designed for internal use only.
Furthermore, it's too opaque to be trusted by a remote user as it only
returns the number of message added to the queue since *the last time
it was called*. The problem here is that the user calling it doesn't
know when it was invoked last. Therefore, they could be getting the
rate of messages added for the last 5 minutes or the last 5
milliseconds. This can lead to inconsistent and misleading results.
There are three main ways for users to track rates of message
production and consumption:
1. Use a metrics plugin. This is the most feature-rich and flexible
way to track broker metrics, although it requires tools (e.g.
Prometheus) to store the metrics and display them (e.g. Grafana).
2. Invoke the getMessageCount() and getMessagesAdded() management
methods and store the returned values along with the time they were
retrieved. A time-series database is a great tool for this job. This is
exactly what tools like Prometheus do. That data can then be used to
create informative graphs, etc. using tools like Grafana. Of course, one
can skip all the tools and just do some simple math to calculate rates
based on the last time the counts were retrieved.
3. Use the broker's message counters. Message counters are the broker's
simple way of providing historical information about the queue. They
provide similar results to the previous solutions, but with less
flexibility since they only track data while the broker is up and
there's not really any good options for graphing.
When performing concurrent user admin actions (e.g. resetUser, addUser,
removeUser on ActiveMQServerControl) when using the
PropertiesLoginModule with reload=true the underlying user and role
properties files can get corrupted.
This commit fixes the issue via the following changes:
- Add synchronization to the management commands
- Add concurrency controls to underlying file access
- Change CLI user commands to use remote methods instead of modifying
the files directly. This avoids potential concurrent changes. This
change forced me to modify the names of some of the commands'
parameters to disambiguate them from connection-related parameters.
In a cluster scenario where non durable subscribers fail over to
backup while another live node forwarding messages to it,
there is a chance that the the live node keeps the old remote
binding for the subs and messages go to those
old remote bindings will result in "binding not found".
The default JAAS security manager doesn't need the address/FQQN for
authorization, but I'm putting it back into the interface because there
are other use cases which *do* need it.
Both authentication and authorization will hit the underlying security
repository (e.g. files, LDAP, etc.). For example, creating a JMS
connection and a consumer will result in 2 hits with the *same*
authentication request. This can cause unwanted (and unnecessary)
resource utilization, especially in the case of networked configuration
like LDAP.
There is already a rudimentary cache for authorization, but it is
cleared *totally* every 10 seconds by default (controlled via the
security-invalidation-interval setting), and it must be populated
initially which still results in duplicate auth requests.
This commit optimizes authentication and authorization via the following
changes:
- Replace our home-grown cache with Google Guava's cache. This provides
simple caching with both time-based and size-based LRU eviction. See more
at https://github.com/google/guava/wiki/CachesExplained. I also thought
about using Caffeine, but we already have a dependency on Guava and the
cache implementions look to be negligibly different for this use-case.
- Add caching for authentication. Both successful and unsuccessful
authentication attempts will be cached to spare the underlying security
repository as much as possible. Authenticated Subjects will be cached
and re-used whenever possible.
- Authorization will used Subjects cached during authentication. If the
required Subject is not in the cache it will be fetched from the
underlying security repo.
- Caching can be disabled by setting the security-invalidation-interval
to 0.
- Cache sizes are configurable.
- Management operations exist to inspect cache sizes at runtime.
This is allowing journal appends to happen in burst
during replication, by batching replication response
into the network at the end of the append burst.
I couldn't reproduce this with a test, but static code analysis led me
to this solution which is similar to the fix done for ARTEMIS-2592 via
e397a17796.
Due to a lack of concurrency protections it's possible to create a
consumer on a closed session. I've not been able to reproduce this with
a test, but I've seen it in the wild. Static code analysis points to a
need for better concurrency controls around closing the session and
creating consumers.
Clebert Suconic
Justin Bertram
Andy Taylor
franz1981
Howard Gao
Domenico Francesco Bruscino
Michal Petrov
gtully
Robbie Gemmell
Martyn Taylor
Christoph Läubrich
AntonRoskvist
Sebastian Thomschke
Markus Meierhofer
Christopher L. Shannon (cshannon)
Petr Kužel
岳豹
Marcos Singermann
Erwin Dondorp
Tomas Hofman
Ryan Yeats
Tadayoshi Sato
Luis De Bello
Emmanuel Hugonnet
Luis Alves
Mikko Uoti
Bernd Gutjahr
Howard Gao
Michael Pearce
brusdev