G1 GC were setup to use an `InitiatingHeapOccupancyPercent` of 75. This
could leave used memory at a very high level for an extended duration,
triggering the real memory circuit breaker even at low activity levels.
The value is a threshold for old generation usage relative to total heap
size and thus it should leave room for the new generation. Default in
G1 is to allow up to 60 percent for new generation and this could mean that the
threshold was effectively at 135% heap usage. GC would still kick in of course and
eventually enough mixed collections would take place such that adaptive adjustment
of IHOP kicks in.
The JVM has adaptive setting of the IHOP, but this does not kick in
until it has sampled a few collections. A newly started, relatively
quiet server with primarily new generation activity could thus
experience heap above 95% frequently for a duration.
The changes here are two-fold:
1. Use 30% default for IHOP (the JVM default of 45 could still mean
105% heap usage threshold and did not fully ensure not to hit the
circuit breaker with low activity)
2. Set G1ReservePercent=25. This is used by the adaptive IHOP mechanism,
meaning old/mixed GC should kick in no later than at 75% heap. This
ensures IHOP stays compatible with the real memory circuit breaker also
after being adjusted by adaptive IHOP.
This PR adds some restrictions around testfixtures to make sure the same service ( as defiend in docker-compose.yml ) is not shared between multiple projects.
Sharing would break running with --parallel.
Projects can still share fixtures as long as each has it;s own service within.
This is still useful to share some of the setup and configuration code of the fixture.
Project now also have to specify a service name when calling useCluster to refer to a specific service.
If this is not the case all services will be claimed and the fixture can't be shared.
For this reason fixtures have to explicitly specify if they are using themselves ( fixture and tests in the same project ).
This commit disables caching of BWC snapshot distributions in the "trunk" (aka master) branch.
Since the previous major release branches move quickly we rarely get cache hits for these
tasks, and the artifacts themselves are very large. This means the overhead here is high and
savings basically zero. We conditionally disable task output caching in this scenario in CI to
avoid excessive build cache overhead as well as causing too much turn in the cache itself which
would lead to lots of cache entry evictions.
This commit teaches the build how to bundle AdoptOpenJDK with our
artifacts, and switches to AdoptOpenJDK as the bundled JDK. We keep the
functionality to also bundle Oracle OpenJDK distributions.
We currently configure io.netty.allocator.numDirectArenas to be 0 in the
jvm erconomics class. This is a config that we always want to set, so it
makes sense to move it to jvm.options.
This adds a pipeline aggregation that calculates the cumulative
cardinality of a field. It does this by iteratively merging in the
HLL sketch from consecutive buckets and emitting the cardinality up
to that point.
This is useful for things like finding the total "new" users that have
visited a website (as opposed to "repeat" visitors).
This is a Basic+ aggregation and adds a new Data Science plugin
to house it and future advanced analytics/data science aggregations.
In the Sys V init scripts, we check for Java. This is not needed, since
the same check happens in elasticsearch-env when starting up. Having
this duplicate check has bitten us in the past, where we made a change
to the logic in elasticsearch-env, but missed updating it here. Since
there is no need for this duplicate check, we remove it from the Sys V
init scripts.
Most of our CLI tools use the Terminal class, which previously did not provide methods for writing to standard output. When all output goes to standard out, there are two basic problems. First, errors and warnings are "swallowed" in pipelines, making it hard for a user to know when something's gone wrong. Second, errors and warnings are intermingled with legitimate output, making it difficult to pass the results of interactive scripts to other tools.
This commit adds a second set of print commands to Terminal for printing to standard error, with errorPrint corresponding to print and errorPrintln corresponding to println. This leaves it to developers to decide which output should go where. It also adjusts existing commands to send errors and warnings to stderr.
Usage is printed to standard output when it's correctly requested (e.g., bin/elasticsearch-keystore --help) but goes to standard error when a command is invoked incorrectly (e.g. bin/elasticsearch-keystore list-with-a-typo | sort).
* Add input and outut tracking of built bwc versions
This PR adds tracking of the bwc versions git has as input and all the
expected files as output.
The effect is that `gradlew` is not called at all when the git has
doesn't change and the version was allready built.
Previusly gradlew would be called for the bwc version and it would have
to configure the project and go trough up to date checks to figure out
that nothing changed.
This helps when working on bwc tests locally needing to run the test
multiple times.
This should also help in CI not re-build bwc versions across different
runs.
* Enable caching of bwc builds
This commit addresses an issue when trying to using Elasticsearch on
systems with Sys V init and the bundled JDK was not being used. Instead,
we were still inadvertently trying to fallback on the path. This commit
removes that fallback as that is against our intentions for 7.x where we
only support the bundled JDK or an explicit JDK via JAVA_HOME.
This commit makes the gitRevision property a lazy loaded value by
returning an Object implementing toString(). The Dockerfile template is
also changed to use groovy templates instead of the mavenfilter hack, so
converting to String will not happen until runtime.
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.
Prior to this PR we always checked out the latest bwc branches and had
an external mechanism to store the bwc versions used for every CI run so
we could both reproduce those builds and run additional tests using the
same combination.
This adds complexities in setting up and maintaining CI and makes it
difficult to set up multi jobs.
This change replaces that mechanism with a time based approach
that looks at the commit date of the current revision and picks the
newest on the bwc branch that's still older than that.
It also makes sure there are no merge commits in this interval.
This new behavior will is ment to be enabled in CI only, for everything
except PR checks that will still use last available bwc revision.
This commit applies a normalization process to environment paths, both
in how they are stored internally, also their settings values. This
normalization is done via two means:
- we make the paths absolute
- we remove redundant name elements from the path (what Java calls
"normalization")
This change ensures that when we compare and refer to these paths within
the system, we are using a common ground. For example, prior to the
change if the data path was relative, we would not compare it correctly
to paths from disk usage. This is because the paths in disk usage were
being made absolute.
The org.label-schema labels on Docker images have been superseded by
pre-defined OCI annotations. However, there is still a lot of tooling in
use that relies on the org.label-schema, so we do not want to drop
them. This commit adds values for the org.opencontainers.image
pre-defined annotation keys. Additionally, we correct an issue with the
label used to represent the license, to use the org.label-schema.license
label. While this label was never accepted into the org.label-schema
specfication (because this specification was superseded, it's not that
it was explicitly rejected) there are containers out there using this
label. In particular, our base image is and so we need to override
otherwise we inherit, and end up mis-reporting the license.
Today our systemd service defaults to a service type of simple. This
means that systemd assumes Elasticsearch is ready as soon as the
ExecStart (bin/elasticsearch) process is forked off. This means that the
service appears ready long before it actually is, so before it is ready
to receive requests. It also means that services that want to depend on
Elasticsearch being ready to start can not as there is not a reliable
mechanism to determine this. This commit changes the service type to
notify. This requires that Elasticsearch sends a notification message
via libsystemd sd_notify method. This commit does that by using JNA to
invoke this native method. Additionally, we use this integration to also
notify systemd when we are stopping.
The field has to be defined in log4j2.properties and should be an
escaped JSON for now (it is a broken JSON at the moment). This should later be refactored into a JSON array
of strings.
* Mute failing test
tracked in #44552
* mute EvilSecurityTests
tracking in #44558
* Fix line endings in ESJsonLayoutTests
* Mute failing ForecastIT test on windows
Tracking in #44609
* mute BasicRenormalizationIT.testDefaultRenormalization
tracked in #44613
* fix mute testDefaultRenormalization
* Increase busyWait timeout windows is slow
* Mute failure unconfigured node name
* mute x-pack internal cluster test windows
tracking #44610
* Mute JvmErgonomicsTests on windows
Tracking #44669
* mute SharedClusterSnapshotRestoreIT testParallelRestoreOperationsFromSingleSnapshot
Tracking #44671
* Mute NodeTests on Windows
Tracking #44256
In https://github.com/elastic/elasticsearch/pull/41913 setting up the
temp dir for ES was moved from the env script to individual cli scripts.
However, moving it to the windows service cli was missed. This commit
restores setting up the temp dir for the windows service control script.
Today when checksumming a plugin zip during plugin install, we read all
of the bytes of the zip into memory at once. When trying to run the
plugin installer on a small heap (say, 64 MiB), this can lead to the
plugin installer running out of memory when checksumming large
plugins. This commit addresses this by reading the plugin bytes in 8 KiB
chunks, thus using a constant amount of memory independent of the size
of the plugin.
This commit adds some default CLI JVM options to control the heap size
and the garbage collector used for the CLI tools. We do this because
otherwise the JVM will default to large initial and max heap sizes based
on the RAM visible to the JVM (which could be all the physical RAM on
the machine if not run in a container-aware JVM). This commit therefore
sets the initial heap size to 4m, the max heap size to 64m, the garbage
collector to the serial collector, and leaves this user-configurable by
honoring ES_JAVA_OPTS last.
This is a refactor to current JSON logging to make it more open for extensions
and support for custom ES log messages used inDeprecationLogger IndexingSlowLog , SearchSLowLog
We want to include x-opaque-id in deprecation logs. The easiest way to have this as an additional JSON field instead of part of the message is to create a custom DeprecatedMessage (extends ESLogMEssage)
These messages are regular log4j messages with a text, but also carry a map of fields which can then populate the log pattern. The logic for this lives in ESJsonLayout and ESMessageFieldConverter.
Similar approach can be used to refactor IndexingSlowLog and SearchSlowLog JSON logs to contain fields previously only present as escaped JSON string in a message field.
closes#41350
backport #41354
This change makes the process of verifying the signature of
official plugins FIPS 140 compliant by defaulting to use the
BouncyCastle FIPS provider and adding a dependency to bcpg-fips
that implement parts of openPGP in a FIPS compliant manner.
In already FIPS 140 enabled environments that use the
BouncyCastle FIPS provider, the bcfips dependency is redundant
but doesn't cause an issue as it will be added only in the classpath
of the cli-tools
This is a backport of #44224
When using gradle run by itself, this uses the default distro with a
basic license and enables security. There is a setup command to create
a elastic-admin user but only when the license is a trial license. Now
that security is available with the basic license, we should always run
this command when using the default distribution.
We initially added `requireDocker` for a way for tasks to say that they
absolutely must have it, like the build docker image tasks.
Projects using the test fixtures plugin are not in this both, as the
intent with these is that they will be skipped if docker and docker-compose
is not available.
Before this change we were lenient, the docker image build would succeed
but produce nothing. The implementation was also confusing as it was not
immediately obvious this was the case due to all the indirection in the
code.
The reason we have this leniency is that when we added the docker image
build, docker was a fairly new requirement for us, and we didn't have
it deployed in CI widely enough nor had CI configured to prefer workers
with docker when possible. We are in a much better position now.
The other reason was other stack teams running `./gradlew assemble`
in their respective CI and the possibility of breaking them if docker is
not installed. We have been advocating for building specific distros for
some time now and I will also send out an additional notice
The PR also removes the use of `requireDocker` from tests that actually
use test fixtures and are ok without it, and fixes a bug in test
fixtures that would cause incorrect configuration and allow some tasks
to run when docker was not available and they shouldn't have.
Closes #42680 and #42829 see also #42719
Enable audit logs in docker by creating console appenders for audit loggers.
also rename field @timestamp to timestamp and add field type with value audit
The docker build contains now two log4j configuration for oss or default versions. The build now allows override the default configuration.
Also changed the format of a timestamp from ISO8601 to include time zone as per this discussion #36833 (comment)
closes#42666
backport#42671
Before this change we would recurse to cache bwc versions.
This proved to be problematic due to the number of steps it was
generating taking too long.
Also this required tricky maintenance to break the recursion for old
branches we don't really care about.
With this change we now cache specific branches only.
Previously we used LoggedExec for running the internal bwc builds.
However, this had bad performance implications as all the output was
buffered into memory, thus we changed back to normal Exec. This commit
adds a `spoolOutput` setting to LoggedExec which can be used for
commands with large amounts of output, and switches the bwc builds to
use this flag.
The elasticsearch-cli helper script does not use the tempdir created by
elasticsearch-env, yet the env script still creates it. This can lead to
lots of temp directories being created when running cli scripts in an
automated fashion. This commit passes a fake tmpdir to the env script to
avoid creation.
closes#34445
This commit adds deletion of the bin directory to postrm cleanup. While
the package's bin files are cleaned up by the package manager, plugins
may have created subdirectories under bin. We already cleanup plugins,
but not the extra bin dirs their installation created.
closes#18109
Java 8 presents the JVM options slightly differently when displaying via
-XX:+PrintFlagsFinal. This commit adapts the JVM options parser for this
possibility.
Relates #42009
This commit removes manual parsing of JVM options when calculating
ergonomics. This is to avoid a situation that we parse values
differently than the JVM would. In fact, we already have a bug along
these lines today. It is possible to start the JVM with the same flag
multiple times on the command line. In this case, the last value
wins. For example, -Xmx1g -Xmx2g would start the JVM with a heap size of
two gigabytes. Our JVM ergonomics ignores this possibility and instead
the first value is winning!
Our strategy to avoid manual parsing of the JVM options is to start the
Java command line parser (without actually starting a JVM) by invoking
java with the same command line flags as presented and request that the
JVM tell us what values it would start with. This ensures that we have
the correct values when making ergonomic decisions.
Moreover, our strategy also is ignoring ES_JAVA_OPTS which could
override the heap size as well leading to incorrect ergonomic
choices. This commit address this issue too.
Henning Andersen
Alpar Torok
Mark Vieira
Jason Tedor
Zachary Tong
Tim Brooks
William Brafford
Ryan Ernst
Ioannis Kakavas
Przemyslaw Gomulka
Chris Koehnke
Ryan Ernst
Vincent Boulaye
Jay Modi