Currently all of our transport protocol decoding and aggregation occurs
in the individual transport modules. This means that each implementation
(test, netty, nio) must implement this logic. Additionally, it means
that the entire message has been read from the network before the server
package receives it.
This commit creates a pipeline in server which can be passed arbitrary
bytes to handle. Internally, the pipeline will decode, decompress, and
aggregate the messages. Additionally, this allows us to run many
megabytes of bytes through the pipeline in tests to ensure that the
logic works.
This work will enable future work:
Circuit breaking or backoff logic based on message type and byte
in the content aggregator.
Sharing bytes with the application layer using the ref counted
releasable network bytes.
Improved network monitoring based specifically on channels.
Finally, this fixes the bug where we do not circuit break on the correct
message size when compression is enabled.
Elasticsearch has a number of different BytesReference implementations.
These implementations can all implement the interface in different ways
with subtly different behavior and performance characteristics. On the
other-hand, the JVM only represents bytes as an array or a direct byte
buffer. This commit deletes the specialized Netty implementations and
moves to using a generic ByteBuffer reference type. This will allow us
to focus on standardizing performance and behave around a smaller number
of implementations that can be used by all components in Elasticsearch.
Now that the FIPS 140 security provider is simply a test dependency
we don't need the thirdPartyAudit exceptions, but plugin-cli and
transport-netty4 do need jarHell disabled as they use the non fips
BouncyCastle security provider as a test dependency too.
This change changes the way to run our test suites in
JVMs configured in FIPS 140 approved mode. It does so by:
- Configuring any given runtime Java in FIPS mode with the bundled
policy and security properties files, setting the system
properties java.security.properties and java.security.policy
with the == operator that overrides the default JVM properties
and policy.
- When runtime java is 11 and higher, using BouncyCastle FIPS
Cryptographic provider and BCJSSE in FIPS mode. These are
used as testRuntime dependencies for unit
tests and internal clusters, and copied (relevant jars)
explicitly to the lib directory for testclusters used in REST tests
- When runtime java is 8, using BouncyCastle FIPS
Cryptographic provider and SunJSSE in FIPS mode.
Running the tests in FIPS 140 approved mode doesn't require an
additional configuration either in CI workers or locally and is
controlled by specifying -Dtests.fips.enabled=true
It is the job of the http server transport to release the request in the handler
but the mock fails to do so since we never override `incomingRequest`.
* Copying the request is not necessary here. We can simply release it once the response has been generated and a lot of `Unpooled` allocations that way
* Relates #32228
* I think the issue that preventet that PR that PR from being merged was solved by #39634 that moved the bulk index marker search to ByteBuf bulk access so the composite buffer shouldn't require many additional bounds checks (I'd argue the bounds checks we add, we save when copying the composite buffer)
* I couldn't neccessarily reproduce much of a speedup from this change, but I could reproduce a very measureable reduction in GC time with e.g. Rally's PMC (4g heap node and bulk requests of size 5k saw a reduction in young GC time by ~10% for me)
Backport of #48849. Update `.editorconfig` to make the Java settings the
default for all files, and then apply a 2-space indent to all `*.gradle`
files. Then reformat all the files.
This commit introduces a consistent, and type-safe manner for handling
global build parameters through out our build logic. Primarily this
replaces the existing usages of extra properties with static accessors.
It also introduces and explicit API for initialization and mutation of
any such parameters, as well as better error handling for uninitialized
or eager access of parameter values.
Closes#42042
BytesReference is currently an abstract class which is extended by
various implementations. This makes it very difficult to use the
delegation pattern. The implication of this is that our releasable
BytesReference is a PagedBytesReference type and cannot be used as a
generic releasable bytes reference that delegates to any reference type.
This commit makes BytesReference an interface and introduces an
AbstractBytesReference for common functionality.
Elasticsearch does not grant Netty reflection access to get Unsafe. The
only mechanism that currently exists to free direct buffers in a timely
manner is to use Unsafe. This leads to the occasional scenario, under
heavy network load, that direct byte buffers can slowly build up without
being freed.
This commit disables Netty direct buffer pooling and moves to a strategy
of using a single thread-local direct buffer for interfacing with sockets.
This will reduce the memory usage from networking. Elasticsearch
currently derives very little value from direct buffer usage (TLS,
compression, Lucene, Elasticsearch handling, etc all use heap bytes). So
this seems like the correct trade-off until that changes.
Uses JDK 11's per-socket configuration of TCP keepalive (supported on Linux and Mac), see
https://bugs.openjdk.java.net/browse/JDK-8194298, and exposes these as transport settings.
By default, these options are disabled for now (i.e. fall-back to OS behavior), but we would like
to explore whether we can enable them by default, in particular to force keepalive configurations
that are better tuned for running ES.
Currently in the transport-nio work we connect and bind channels on the
a thread before the channel is registered with a selector. Additionally,
it is at this point that we set all the socket options. This commit
moves these operations onto the event-loop after the channel has been
registered with a selector. It attempts to set the socket options for a
non-server channel at registration time. If that fails, it will attempt
to set the options after the channel is connected. This should fix
#41071.
* Stop Passing Around REST Request in Multiple Spots
* Motivated by #44564
* We are currently passing the REST request object around to a large number of places. This works fine since we simply copy the full request content before we handle the rest itself which is needlessly hard on GC and heap.
* This PR removes a number of spots where the request is passed around needlessly. There are many more spots to optimize in follow-ups to this, but this one would already enable bypassing the request copying for some error paths in a follow up.
Today we have an annotation for controlling logging levels in
tests. This annotation serves two purposes, one is to control the
logging level used in tests, when such control is needed to impact and
assert the behavior of loggers in tests. The other use is when a test is
failing and additional logging is needed. This commit separates these
two concerns into separate annotations.
The primary motivation for this is that we have a history of leaving
behind the annotation for the purpose of investigating test failures
long after the test failure is resolved. The accumulation of these stale
logging annotations has led to excessive disk consumption. Having
recently cleaned this up, we would like to avoid falling into this state
again. To do this, we are adding a link to the test failure under
investigation to the annotation when used for the purpose of
investigating test failures. We will add tooling to inspect these
annotations, in the same way that we have tooling on awaits fix
annotations. This will enable us to report on the use of these
annotations, and report when stale uses of the annotation exist.
This commit moves the config that stores Cors options into the server
package. Currently both nio and netty modules must have a copy of this
config. Moving it into server allows one copy and the tests to be in a
common location.
Registering a channel with a selector is a required operation for the
channel to be handled properly. Currently, we mix the registeration with
other setup operations (ip filtering, SSL initiation, etc). However, a
fail to register is fatal. This PR modifies how registeration occurs to
immediately close the channel if it fails.
There are still two clear loopholes for how a user can interact with a
channel even if registration fails. 1. through the exception handler.
2. through the channel accepted callback. These can perhaps be improved
in the future. For now, this PR prevents writes from proceeding if the
channel is not registered.
Simplifies AbstractSimpleTransportTestCase to use JVM-local ports and also adds an assertion so
that cases like #44134 can be more easily debugged. The likely reason for that one is that a test,
which was repeated again and again while always spawning a fresh Gradle worker (due to Gradle
daemon) kept increasing Gradle worker IDs, causing an overflow at some point.
Currently nio implements ip filtering at the channel context level. This
is kind of a hack as the application logic should be implemented at the
handler level. This commit moves the ip filtering into a channel
handler. This requires adding an indicator to the channel handler to
show when a channel should be closed.
We had this as a dependency for legacy dependencies that still needed
the Log4j 1.2 API. This appears to no longer be necessary, so this
commit removes this artifact as a dependency.
To remove this dependency, we had to fix a few places where we were
accidentally relying on Log4j 1.2 instead of Log4j 2 (easy to do, since
both APIs were on the compile-time classpath).
Finally, we can remove our custom Netty logger factory. This was needed
when we were on Log4j 1.2 and handled logging in our own unique
way. When we migrated to Log4j 2 we could have dropped this
dependency. However, even then Netty would still pick up Log4j 1.2 since
it was on the classpath, thus the advantage to removing this as a
dependency now.
This is related to #27260. Currently we have a single read buffer that
is no larger than a single TLS packet. This prevents us from reading
multiple TLS packets in a single socket read call. This commit modifies
our TLS work to support reading similar to the plaintext case. The data
will be copied to a (potentially) recycled TLS packet-sized buffer for
interaction with the SSLEngine.
This fixes#41794. Currently the read timeout test queues up responses
in the netty pipeline. These responses are immediately returned in the
write call, but they are not released. This commit releases the
responses. This will cause the leak detector to quit throwing
exceptions.
This is related to #27260. Currently there is a setting
http.read_timeout that allows users to define a read timeout for the
http transport. This commit implements support for this functionality
with the transport-nio plugin. The behavior here is that a repeating
task will be scheduled for the interval defined. If there have been
no requests received since the last run and there are no inflight
requests, the channel will be closed.
This is related to #27260. Currently for the SSLDriver we allocate a
dedicated network write buffer and encrypt the data into that buffer one
buffer at a time. This requires constantly switching between encrypting
and flushing. This commit adds a dedicated outbound buffer for SSL
operations that will internally allocate new packet sized buffers as
they are need (for writing encrypted data). This allows us to totally
encrypt an operation before writing it to the network. Eventually it can
be hooked up to buffer recycling.
This commit also backports the following commit:
Handle WRAP ops during SSL read
It is possible that a WRAP operation can occur while decrypting
handshake data in TLS 1.3. The SSLDriver does not currently handle this
well as it does not have access to the outbound buffer during read call.
This commit moves the buffer into the Driver to fix this issue. Data
wrapped during a read call will be queued for writing after the read
call is complete.
This is related to #27260. Currently for the SSLDriver we allocate a
dedicated network write buffer and encrypt the data into that buffer one
buffer at a time. This requires constantly switching between encrypting
and flushing. This commit adds a dedicated outbound buffer for SSL
operations that will internally allocate new packet sized buffers as
they are need (for writing encrypted data). This allows us to totally
encrypt an operation before writing it to the network. Eventually it can
be hooked up to buffer recycling.
The xlint exclusions of the following plugins were removed:
* ingest-attachment.
* mapper-size.
* transport-nio. Removing the -try exclusion required some work, because
the NettyAdaptor implements AutoCloseable and NettyAdaptor#close() method
could throw an InterruptedException (ChannelFuture#await() and a generic
Exception is re-thrown, which maybe an ChannelFuture). The easiest way
around this to me seemed that NettyAdaptor should not implement AutoCloseable,
because it is not directly used in a try-with-resources statement.
Relates to #40366
This commit adds an InboundHandler to handle inbound message processing.
With this commit, this code is moved out of the TcpTransport.
Additionally, finer grained unit tests are added to ensure that the
inbound processing works as expected
Currently there are some components of message serializer and sending
that still occur in TcpTransport. This commit makes it possible to
send a message without the TcpTransport by moving all of the remaining
application logic to the OutboundHandler. Additionally, it adds unit
tests to ensure that this logic works as expected.
Currently, we maintain a transport name ("mock-nio", "nio", "netty")
that is passed to a `TcpTransportChannel` when a request is received.
The value of this name is to associate with the task when we register a
task with the task manager. However, it is only possible to run ES with
one transport, so having an implementation specific name is unnecessary.
This commit removes the name and replaces it with the generic
"transport".
This change marks the transport-nio plugin as having a client jar. The
nio transport can be used from a transport client and the
x-pack-transport artifact depends on the transport-nio jar but this jar
is not published. This change marks the transport-nio project as having
a client jar so that the jar may be published in the same way that we
publish the netty4 transport artifact.
The change to actually publish the jar will be handled separately as an
update to the release manager.
* Optimize Bulk Message Parsing and Message Length Parsing
* findNextMarker took almost 1ms per invocation during the PMC rally track
* Fixed to be about an order of magnitude faster by using Netty's bulk `ByteBuf` search
* It is unnecessary to instantiate an object (the input stream wrapper) and throw it away, just to read the `int` length from the message bytes
* Fixed by adding bulk `int` read to BytesReference
* Un-mute and fix BuildExamplePluginsIT
There doesn't seem to be anything wrong with the test iteself.
I think the failure were CI performance related, but while it was muted,
some failures managed to sneak in.
Closes#38784
* PR review
This commit introduces a NetworkMessage class. This class has two
subclasses - InboundMessage and OutboundMessage. These messages can
be serialized and deserialized independent of the transport. This allows
more granular testing. Additionally, the serialization mechanism is now
a simple Supplier. This builds the framework to eventually move the
serialization of transport messages to the network thread. This is the
one serialization component that is not currently performed on the
network thread (transport deserialization and http serialization and
deserialization are all on the network thread).
Currently we create dedicated network threads for both the http and
transport implementations. Since these these threads should never
perform blocking operations, these threads could be shared. This commit
modifies the nio-transport to have 0 http workers be default. If the
default configs are used, this will cause the http transport to be run
on the transport worker threads. The http worker setting will still exist
in case the user would like to configure dedicated workers. Additionally,
this commmit deletes dedicated acceptor threads. We have never had these
for the netty transport and they can be added back if a need is
determined in the future.
This commit updates our transport settings for 7.0. It generally takes a
few approaches. First, for normal transport settings, it usestransport.
instead of transport.tcp. Second, it uses transport.tcp, http.tcp,
or network.tcp for all settings that are proxies for OS level socket
settings. Third, it marks the network.tcp.connect_timeout setting for
removal. Network service level settings are only settings that apply to
both the http and transport modules. There is no connect timeout in
http. Fourth, it moves all the transport settings to a single class
TransportSettings similar to the HttpTransportSettings class.
This commit does not actually remove any settings. It just adds the new
renamed settings and adds todos for settings that will be deprecated.
Currently our handshake requests do not include a version. This is
unfortunate as we cannot rely on the stream version since it is not the
sending node's version. Instead it is the minimum compatibility version.
The handshake request is currently empty and we do nothing with it. This
should allow us to add data to the request without breaking backwards
compatibility.
This commit adds the version to the handshake request. Additionally, it
allows "future data" to be added to the request. This allows nodes to craft
a version compatible response. And will properly handle additional data in
future handshake requests. The proper handling of "future data" is useful
as this is the only request where we do not know the other node's version.
Finally, it renames the TcpTransportHandshaker to
TransportHandshaker.
This commit modifies BigArrays to take a circuit breaker name and
the circuit breaking service. The default instance of BigArrays that
is passed around everywhere always uses the request breaker. At the
network level, we want to be using the inflight request breaker. So this
change will allow that.
Additionally, as this change moves away from a single instance of
BigArrays, the class is modified to not be a Releasable anymore.
Releasing big arrays was always dispatching to the PageCacheRecycler,
so this change makes the PageCacheRecycler the class that needs to be
managed and torn-down.
Finally, this commit closes#31435 be making the serialization of
transport messages use the inflight request breaker. With this change,
we no longer push the global BigArrays instnace to the network level.
This commit upgrades netty. This will close#35360. Netty started
throwing an IllegalArgumentException if a CompositeByteBuf is
created with < 2 components. Netty4Utils was updated to reflect this
change.
This is related to #34405 and a follow-up to #34753. It makes a number
of changes to our current keepalive pings.
The ping interval configuration is moved to the ConnectionProfile.
The server channel now responds to pings. This makes the keepalive
pings bidirectional.
On the client-side, the pings can now be optimized away. What this
means is that if the channel has received a message or sent a message
since the last pinging round, the ping is not sent for this round.
This commit removes the dedicated `setSoLinger` method. This simplifies
the `TcpChannel` interface. This method has very little effect as the
SO_LINGER is not set prior to the channels being closed in the abstract
transport test case. We still will set SO_LINGER on the
`MockNioTransport`. However we can do this manually.
This commit is related to #32517. It allows an "sni_server_name"
attribute on a DiscoveryNode to be propagated to the server using
the TLS SNI extentsion. Prior to this commit, this functionality
was only support for the netty transport. This commit adds this
functionality to the security nio transport.
This is related to #34483. It introduces a namespaced setting for
compression that allows users to configure compression on a per remote
cluster basis. The transport.tcp.compress remains as a fallback
setting. If transport.tcp.compress is set to true, then all requests
and responses are compressed. If it is set to false, only requests to
clusters based on the cluster.remote.cluster_name.transport.compress
setting are compressed. However, after this change regardless of any
local settings, responses will be compressed if the request that is
received was compressed.
This is related to #29023. Additionally at other points we have
discussed a preference for removing the need to unnecessarily block
threads for opening new node connections. This commit lays the groudwork
for this by opening connections asynchronously at the transport level.
We still block, however, this work will make it possible to eventually
remove all blocking on new connections out of the TransportService
and Transport.
Currently we create a new netty event loop group for client connections
and all server profiles. Each new group creates new threads for io
processing. This means 2 * num of processors new threads for each group.
A single group should be able to handle all io processing (for the
transports). This also brings the netty module inline with what we do
for nio.
Additionally, this PR renames the worker threads to be the same for
netty and nio.
This is related to #30876. The AbstractSimpleTransportTestCase initiates
many tcp connections. There are normally over 1,000 connections in
TIME_WAIT at the end of the test. This is because every test opens at
least two different transports that connect to each other with 13
channel connection profiles. This commit modifies the default
connection profile used by this test to 6. One connection for each
type, except for REG which gets 2 connections.
Currently a bad regex in CORS settings throws a PatternSyntaxException, which
then bubbles up through the bootstrap code, meaning users have to parse a
stack trace to work out where the problem is. We should instead catch this
exception and rethrow with a more useful error message.
This commit introduces an AbstractSimpleSecurityTransportTestCase for
security transports. This classes provides transport tests that are
specific for security transports. Additionally, it fixes the tests referenced in
#33285.
Drops `Settings` from some of the methods to lookup loggers and
deprecates another logger lookup that takes `Settings` because
`Settings` is no longer required to build a logger.
Historically we have had a ESLoggingHandler in the netty module that
logs low-level connection operations. This class just extends the netty
logging handler with some (broken) message deserialization. This commit
fixes this message serialization and moves the class to server.
This new logger logs inbound and outbound messages. Eventually, we
should move other event logging to this class (connect, close, flush).
That way we will have consistent logging regards of which transport is
loaded.
Resolves#27306 on master. Older branches will need a different fix.
- third party audit detects jar hell with JDK so we disable it
- jdk non portable in forbiddenapis detects classes being used from the
JDK ( for fips ) that are not portable, this is intended so we don't
scan for it on fips.
- different exclusion rules for third party audit on fips
Closes#33179
In our Netty layer we have had to take extra precautions against Netty
catching throwables which prevents them from reaching the uncaught
exception handler. This code has taken on additional uses in NIO layer
and now in the scheduler engine because there are other components in
stack traces that could catch throwables and suppress them from reaching
the uncaught exception handler. This commit is a simple cleanup of the
iterative evolution of this code to refactor all uses into a single
method in ExceptionsHelper.
This is related to #32517. This commit passes the DiscoveryNode to the
initiateChannel method for different Transport implementation. This
will allow additional attributes (besides just the socket address) to be
used when opening channels.
This is a followup to #31886. After that commit the
TransportConnectionListener had to be propogated to both the
Transport and the ConnectionManager. This commit moves that listener
to completely live in the ConnectionManager. The request and response
related methods are moved to a TransportMessageListener. That listener
continues to live in the Transport class.
This is related to #27260. This commit replaces the netty driven http
client (Netty4HttpClient) with one that is driven by (NioHttpClient).
This client exists in the test package and is used for making http
requests.
* Upgrade to `4.1.28` since the problem reported in #32487 is a bug in Netty itself (see https://github.com/netty/netty/issues/7337)
* Fixed other leaks in test code that now showed up due to fixes improvements in leak reporting in the newer version
* Needed to extend permissions for netty common package because it now sets a classloader at runtime after changes in 63bae0956a
* Adjusted forbidden APIs check accordingly
* Closes#32487
There are two scenarios where a http request could terminate in the cors
handler. If that occurs, the requests need to be released. This commit
releases those requests.
* TESTS: Fix Buf Leaks in HttpReadWriteHandlerTests
* Release all ref counted things that weren't getting properly released
* Mannually force channel promise to be completed because mock channel doesn't do it and it prevents one `release` call in `io.netty.channel.ChannelOutboundHandlerAdapter#write` from firing
This is related to #27260. It adds the SecurityNioHttpServerTransport
to the security plugin. It randomly uses the nio http transport in
security integration tests.
Currently we set local addresses on the creation time of a NioChannel.
However, this may return null as the local address may not have been
set yet. An example is the local address has not been set on a client
channel as the connection process is not yet complete.
This PR modifies the getter to set the local field if it is currently null.
Historically in TcpTransport server channels were represented by the
same channel interface as socket channels. This was necessary as
TcpTransport was parameterized by the channel type. This commit
introduces TcpServerChannel and HttpServerChannel classes. Additionally,
it adds the implementations for the various transports. This allows
server channels to have unique functionality and not implement the
methods they do not support (such as send and getRemoteAddress).
Additionally, with the introduction of HttpServerChannel this commit
extracts some of the storing and closing channel work to the abstract
http server transport.
This is a general cleanup of channels and exception handling in http.
This commit introduces a CloseableChannel that is a superclass of
TcpChannel and HttpChannel. This allows us to unify the closing logic
between tcp and http transports. Additionally, the normal http channels
are extracted to the abstract server transport.
Finally, this commit (mostly) unifies the exception handling between nio
and netty4 http server transports.
This is related to #28898. This PR implements pooling of bytes arrays
when reading from the wire in the http server transport. In order to do
this, we must integrate with netty reference counting. That manner in
which this PR implements this is making Pages in InboundChannelBuffer
reference counted. When we accessing the underlying page to pass to
netty, we retain the page. When netty releases its bytebuf, it releases
the underlying pages we have passed to it.
Currently we maintain a compatibility map of http status codes in both
the netty4 and nio modules. These maps convert a RestStatus to a netty
HttpResponseStatus. However, as these fundamentally represent integers,
we can just use the netty valueOf method to convert a RestStatus to a
HttpResponseStatus.
This is related to #28898. With the addition of the http nio transport,
we now have two different modules that provide http transports.
Currently most of the http logic lives at the module level. However,
some of this logic can live in server. In particular, some of the
setting of headers, cors, and pipelining. This commit begins this moving
in that direction by introducing lower level abstraction (HttpChannel,
HttpRequest, and HttpResonse) that is implemented by the modules. The
higher level rest request and rest channel work can live entirely in
server.
Currently the http pipelining handlers seem to support chunked http
content. However, this does not make sense. There is a content
aggregator in the pipeline before the pipelining handler. This means the
pipelining handler should only see full http messages. Additionally, the
request handler immediately after the pipelining handler only supports
full messages.
This commit modifies both nio and netty4 pipelining handlers to assert
that an inbound message is a full http message. Additionally it removes
the tests for chunked content.
This commit upgrades us to Netty 4.1.25. This upgrade is more
challenging than past upgrades, all because of a new object cleaner
thread that they have added. This thread requires an additional security
permission (set context class loader, needed to avoid leaks in certain
scenarios). Additionally, there is not a clean way to shutdown this
thread which means that the thread can fail thread leak control during
tests. As such, we have to filter this thread from thread leak control.
This is related to #27260 and #28898. This commit adds the transport-nio
plugin as a random option when running the http smoke tests. As part of
this PR, I identified an issue where cors support was not properly
enabled causing these tests to fail when using transport-nio. This
commit also fixes that issue.
This is related to #28898. This commit adds the acceptor thread name to
the method checking if this thread is a transport thread. Additionally,
it modifies the nio http transport to use the same worker name as the
netty4 http server transport.
This is related to #27260. This commit combines the AcceptingSelector
and SocketSelector classes into a single NioSelector. This change
allows the same selector to handle both server and socket channels. This
is valuable as we do not necessarily want a dedicated thread running for
accepting channels.
With this change, this commit removes the configuration for dedicated
accepting selectors for the normal transport class. The accepting
workload for new node connections is likely low, meaning that there is
no need to dedicate a thread to this process.
This is related to #28898. This commit adds cors support to the nio http
transport. Most of the work is copied directly from the netty module
implementation. Additionally, this commit adds tests for the nio http
channel.
Tim Brooks
Ioannis Kakavas
Armin Braun
Maria Ralli
Rory Hunter
Mark Vieira
Yannick Welsch
Jason Tedor
Alan Woodward
Martijn van Groningen
Jay Modi
Alpar Torok
Gordon Brown
Pratik Sanglikar
Alan Woodward
Nik Everett
Or Bin
Christoph Büscher
Igor Motov
Tanguy Leroux