The current shard follow mechanism is complex and does not give us easy ways the have visibility into the system (e.g. why we are falling behind).
The main reason why it is complex is because the current design is highly asynchronous. Also in the current model it is hard to apply backpressure
other than reducing the concurrent reads from the leader shard.
This PR has the following changes:
* Rewrote the shard follow task to coordinate the shard follow mechanism between a leader and follow shard in a single threaded manner.
This allows for better unit testing and makes it easier to add stats.
* All write operations read from the shard changes api should be added to a buffer instead of directly sending it to the bulk shard operations api.
This allows to apply backpressure. In this PR there is a limit that controls how many write ops are allowed in the buffer after which no new reads
will be performed until the number of ops is below that limit.
* The shard changes api includes the current global checkpoint on the leader shard copy. This allows reading to be a more self sufficient process;
instead of relying on a background thread to fetch the leader shard's global checkpoint.
* Reading write operations from the leader shard (via shard changes api) is a separate step then writing the write operations (via bulk shards operations api).
Whereas before a read would immediately result into a write.
* The bulk shard operations api returns the local checkpoint on the follow primary shard, to keep the shard follow task up to date with what has been written.
* Moved the shard follow logic that was previously in ShardFollowTasksExecutor to ShardFollowNodeTask.
* Moved over the changes from #31242 to make shard follow mechanism resilient from node and shard failures.
Relates to #30086
* master:
[ML] Fix master node deadlock during ML daily maintenance (#31836)
Build: Switch integ-test-zip to OSS-only (#31866)
SQL: Remove restriction for single column grouping (#31818)
Build: Fix detection of Eclipse Compiler Server (#31838)
Docs: Inconsistency between description and example (#31858)
Re-enable bwc tests now that #29538 has been backported and 6.x intake build succeeded.
QA: build improvements related to SQL projects (#31862)
[Docs] Add clarification to analysis example (#31826)
Check timeZone() argument in AbstractSqlQueryRequest (#31822)
SQL: Fix incorrect HAVING equality (#31820)
Smaller aesthetic fixes to InternalTestCluster (#31831)
[Docs] Clarify accepted sort case (#31605)
Temporarily disable bwc test in order to backport #29538
Remove obsolete parameters from analyze rest spec (#31795)
[Docs] Fix wrong link in Korean analyzer docs (#31815)
Fix profiling of ordered terms aggs (#31814)
Properly mute test involving JDK11 closes#31739
Do not return all indices if a specific alias is requested via get aliases api. (#29538)
Get snapshot rest client cleanups (#31740)
Docs: Explain _bulk?refresh shard targeting
Fix handling of points_only with term strategy in geo_shape (#31766)
For historical reasons SQL restricts GROUP BY to only one field.
This commit removes the restriction and improves the test suite with
multi group by tests.
Close#31793
If a get alias api call requests a specific alias pattern then
indices not having any matching aliases should not be included in the response.
This is a second attempt to fix this (first attempt was #28294).
The reason that the first attempt was reverted is because when xpack
security is enabled then index expression (like * or _all) are resolved
prior to when a request is processed in the get aliases transport action,
then `MetaData#findAliases` can't know whether requested all where
requested since it was already expanded in concrete alias names. This
change replaces aliases(...) replaceAliases(...) method on AliasesRequests
class and leave the aliases(...) method on subclasses. So there is a distinction
between when xpack security replaces aliases and a user setting aliases via
the transport or high level http client.
Closes#27763
* master:
REST high-level client: add get index API (#31703)
SQL: Allow long literals (#31777)
SQL: Fix incorrect message for aliases (#31792)
Test: Do not remove xpack templates when cleaning (#31642)
Reduce more raw types warnings (#31780)
Add unreleased version 6.3.2
Scripting: Remove support for deprecated StoredScript contexts (#31394)
[ML][TEST] Use java 11 valid time format in DataDescriptionTests (#31817)
[ML] Don't treat stale FAILED jobs as OPENING in job allocation (#31800)
[ML] Fix calendar and filter updates from non-master nodes (#31804)
Fix license header generation on Windows (#31790)
mark RollupIT.testTwoJobsStartStopDeleteOne as AwaitsFix
mark SearchAsyncActionTests.testFanOutAndCollect as AwaitsFix
Correct exclusion of test on JDK 11
Fix doclint jdk 11
Add JDK11 support and enable in CI (#31644)
Watcher: Fix check for currently executed watches (#31137)
Watcher: Ensure correct method is used to read secure settings (#31753)
SQL: Update CLI logo
Removes support for storing scripts without the usual json around the
script. So You can no longer do:
```
POST _scripts/<templatename>
{
"query": {
"match": {
"title": "{{query_string}}"
}
}
}
```
and must instead do:
```
POST _scripts/<templatename>
{
"script": {
"lang": "mustache",
"source": {
"query": {
"match": {
"title": "{{query_string}}"
}
}
}
}
}
```
This improves error reporting when you attempt to store a script but don't
quite get the syntax right. Before, there was a good chance that we'd
think of it as a "raw" template and just store it. Now we won't do that.
Nice.
It seems that java 11 tightened some validations with regard to
time formats. The random instance creator was setting an odd
time format to the data description which is invalid when run
with java 11. This commit changes it to a valid format.
Job persistent tasks with stale allocation IDs used to always be
considered as OPENING jobs in the ML job node allocation decision.
However, FAILED jobs are not relocated to other nodes, which leads
to them blocking up the nodes they failed on after node restarts.
FAILED jobs should not restrict how many other jobs can open on a
node, regardless of whether they are stale or not.
Closes#31794
Job updates or changes to calendars or filters may
result into updating the job process if it has been
running. To preserve the order of updates, process
updates are queued through the UpdateJobProcessNotifier
which is only running on the master node. All actions
performing such updates must run on the master node.
However, the CRUD actions for calendars and filters
are not master node actions. They have been submitting
the updates to the UpdateJobProcessNotifier even though
it might have not been running (given the action was
run on a non-master node). When that happens, the update
never reaches the process.
This commit fixes this problem by ensuring the notifier
runs on all nodes and by ensuring the process update action
gets the resources again before updating the process
(instead of having those resources passed in the request).
This ensures that even if the order of the updates
gets messed up, the latest update will read the latest
state of those resource and the process will get back
in sync.
This leaves us with 2 types of updates:
1. updates to the job config should happen on the master
node. This is because we cannot refetch the entire job
and update it. We need to know the parts that have been changed.
2. updates to resources the job uses. Those can be handled
on non-master nodes but they should be re-fetched by the
update process action.
Closes#31803
* Upgrade bouncycastle
Required to fix
`bcprov-jdk15on-1.55.jar; invalid manifest format `
on jdk 11
* Downgrade bouncycastle to avoid invalid manifest
* Add checksum for new jars
* Update tika permissions for jdk 11
* Mute test failing on jdk 11
* Add JDK11 to CI
* Thread#stop(Throwable) was removed
http://mail.openjdk.java.net/pipermail/core-libs-dev/2018-June/053536.html
* Disable failing tests #31456
* Temprorarily disable doc tests
To see if there are other failures on JDK11
* Only blacklist specific doc tests
* Disable only failing tests in ingest attachment plugin
* Mute failing HDFS tests #31498
* Mute failing lang-painless tests #31500
* Fix backwards compatability builds
Fix JAVA version to 10 for ES 6.3
* Add 6.x to bwx -> java10
* Prefix out and err from buildBwcVersion for readability
```
> Task :distribution:bwc:next-bugfix-snapshot:buildBwcVersion
[bwc] :buildSrc:compileJava
[bwc] WARNING: An illegal reflective access operation has occurred
[bwc] WARNING: Illegal reflective access by org.codehaus.groovy.reflection.CachedClass (file:/home/alpar/.gradle/wrapper/dists/gradle-4.5-all/cg9lyzfg3iwv6fa00os9gcgj4/gradle-4.5/lib/groovy-all-2.4.12.jar) to method java.lang.Object.finalize()
[bwc] WARNING: Please consider reporting this to the maintainers of org.codehaus.groovy.reflection.CachedClass
[bwc] WARNING: Use --illegal-access=warn to enable warnings of further illegal reflective access operations
[bwc] WARNING: All illegal access operations will be denied in a future release
[bwc] :buildSrc:compileGroovy
[bwc] :buildSrc:writeVersionProperties
[bwc] :buildSrc:processResources
[bwc] :buildSrc:classes
[bwc] :buildSrc:jar
```
* Also set RUNTIME_JAVA_HOME for bwcBuild
So that we can make sure it's not too new for the build to understand.
* Align bouncycastle dependency
* fix painles array tets
closes#31500
* Update jar checksums
* Keep 8/10 runtime/compile untill consensus builds on 11
* Only skip failing tests if running on Java 11
* Failures are dependent of compile java version not runtime
* Condition doc test exceptions on compiler java version as well
* Disable hdfs tests based on runtime java
* Set runtime java to minimum supported for bwc
* PR review
* Add comment with ticket for forbidden apis
The ack watch action has a check for currently executed watches, to make
sure that currently running watches cannot be acknowledged. This check
only checked on the coordinating node for watches being executed, but should
have checked the whole cluster using a WatcherStatsRequest, which is
being switched to in this commit.
As SecureSetting is extended from Setting, you can easily accidentally
use `SecureSetting.simpleString()` to read a secure setting instead of
`SecureSetting.secureString()`. This commit changes this behaviour in
some watcher notification services.
* master:
[ML] Rate limit established model memory updates (#31768)
[Docs] Correct default window_size (#31582)
S3 fixture should report 404 on unknown bucket (#31782)
Detach Transport from TransportService (#31727)
[ML] Limit ML filter items to 10K (#31731)
[ML] Return statistics about forecasts as part of the jobsstats and usage API (#31647)
Fixture for Minio testing (#31688)
[DOCS] Add missing get mappings docs to HLRC (#31765)
[DOCS] Starting Elasticsearch (#31701)
Painless: Complete Removal of Painless Type (#31699)
Fix not waiting for Netty ThreadDeathWatcher in IT (#31758)
Consolidate watcher setting update registration (#31762)
Build: re-enabled bwc (#31769)
ingest: Introduction of a bytes processor (#31733)
Fix coerce validation_method in GeoBoundingBoxQueryBuilder (#31747)
Add analyze API to high-level rest client (#31577)
[DOCS] Typos
DOC: Add examples to the SQL docs (#31633)
Add support for AWS session tokens (#30414)
Watcher: Reenable start/stop yaml tests (#31754)
Implemented XContent serialisation for GetIndexResponse (#31675)
JDBC: Fix stackoverflow on getObject and timestamp conversion (#31735)
resolveHasher defaults to NOOP (#31723)
Account for XContent overhead in in-flight breaker
Split CircuitBreaker-related tests (#31659)
Add write*Blob option to replace existing blob (#31729)
Painless: Add Context Docs (#31190)
Watcher: Fix chain input toXcontent serialization (#31721)
Docs: Match the examples in the description (#31710)
rest-high-level: added get cluster settings (#31706)
[Docs] Correct typos (#31720)
Clean up double semicolon code typos (#31687)
[DOCS] Check for Windows and *nix file paths (#31648)
[ML] Validate ML filter_id (#31535)
Revert long lines
Fix TransportChangePasswordActionTests
There is at most one model size stats document per bucket, but
during lookback a job can churn through many buckets very quickly.
This can lead to many cluster state updates if established model
memory needs to be updated for a given model size stats document.
This change rate limits established model memory updates to one
per job per 5 seconds. This is done by scheduling the updates 5
seconds in the future, but replacing the value to be written if
another model size stats document is received during the waiting
period. Updating the values in arrears like this means that the
last value received will be the one associated with the job in the
long term, whereas alternative approaches such as not updating the
value if a new value was close to the old value would not.
Today TransportService is tightly coupled with Transport since it
requires an instance of TransportService in order to receive responses
and send requests. This is mainly due to the Request and Response handlers
being maintained in TransportService but also because of the lack of a proper
callback interface.
This change moves request handler registry and response handler registration into
Transport and adds all necessary methods to `TransportConnectionListener` in order
to remove the `TransportService` dependency from `Transport`
Transport now accepts one or more `TransportConnectionListener` instances that are
executed sequentially in a blocking fashion.
Add hard limit to the number of items
a filter may have. This serves to protect
from excessive overhead due to the filters
taking too much memory or lookups becoming
too expensive.
This change adds stats about forecasts, to the jobstats api as well as xpack/_usage. The following
information is collected:
_xpack/ml/anomaly_detectors/{jobid|_all}/_stats:
- total number of forecasts
- memory statistics (mean/min/max)
- runtime statistics
- record statistics
- counts by status
_xpack/usage
- collected by job status as well as overall (_all):
- total number of forecasts
- number of jobs that have at least 1 forecast
- memory, runtime, record statistics
- counts by status
Fixes#31395
Previously the call to register a listener for settings updates was in
each individual service, rather than in the notification service
itself. This change ensures that each child of the notification service
gets registered with the settings update consumer.
Significantly improve the example snippets in the documentation.
The examples are part of the test suite and checked nightly.
To help readability, the existing dataset was extended (test_emp renamed
to emp plus library).
Improve output of JDBC tests to be consistent with the CLI
Add lenient flag to JDBC asserts to allow type widening (a long is
equivalent to a integer as long as the value is the same).
* Default resolveFromHash to Hasher.NOOP
This changes the default behavior when resolving the hashing
algorithm from unrecognised hash strings, which was introduced in
#31234
A hash string that doesn't start with an algorithm identifier can
either be a malformed/corrupted hash or a plaintext password when
Hasher.NOOP is used(against warnings).
Do not make assumptions about which of the two is true for such
strings and default to Hasher.NOOP. Hash verification will subsequently
fail for malformed hashes.
Finally, do not log the potentially malformed hash as this can very
well be a plaintext password.
Resolves#31697
Reverts 58cf95a06f
The xcontent parameters were not passed to the xcontent serialization
of the chain input for each chain. This could lead to wrongly stored
watches, which did not contain passwords but only their redacted counterparts, when an input inside of a chain input contained a password.
testIncorrectPasswordHashingAlgorithm is based on the assumption
that the algorithm selected for the change password request is
different than the one selected for the NativeUsersStore.
pbkdf2_10000 is the same as pbkdf2 since 10000 is the default cost
factor for pbkdf2 and thus should not be used as an option for the
passwordHashingSettings.
Also make sure that the same algorithm is used for settings and
change password requests in other tests for consistency, even if
we expect to not reach the code where the algorithm is checked for
now.
Resolves#31696
Reverts 1c4f480794
* master:
Mute 'Test typed keys parameter for suggesters' as we await a fix.
Build test: Thread linger
Fix gradle4.8 deprecation warnings (#31654)
Mute FileRealmTests#testAuthenticateCaching with an @AwaitsFix.
Mute TransportChangePasswordActionTests#testIncorrectPasswordHashingAlgorithm with an @AwaitsFix.
Build: Fix naming conventions task (#31681)
Introduce a Hashing Processor (#31087)
* remove explicit wrapper task
It's created by Gradle and triggers a deprecation warning
Simplify configuration
* Upgrade shadow plugin to get rid of Gradle deprecation
* Move compile configuration to base plugin
Solves Gradle deprecation warning from earlier Gradle versions
* Enable stable publishing in the Gradle build
* Replace usage of deprecated property
* bump Gradle version in build compare
It is useful to have a processor similar to
logstash-filter-fingerprint
in Elasticsearch. A processor that leverages a variety of hashing algorithms
to create cryptographically-secure one-way hashes of values in documents.
This processor introduces a pbkdf2hmac hashing scheme to fields in documents
for indexing
* master:
Do not check for object existence when deleting repository index files (#31680)
Remove extra check for object existence in repository-gcs read object (#31661)
Support multiple system store types (#31650)
[Test] Clean up some repository-s3 tests (#31601)
[Docs] Use capital letters in section headings (#31678)
[DOCS] Add PQL language Plugin (#31237)
Merge AzureStorageService and AzureStorageServiceImpl and clean up tests (#31607)
TEST: Fix test task invocation (#31657)
Revert "[TEST] Mute failing tests in NativeRealmInteg and ReservedRealmInteg"
Fix RealmInteg test failures
Extend allowed characters for grok field names (#21745) (#31653)
[DOCS] Fix licensing API details (#31667)
[TEST] Mute failing tests in NativeRealmInteg and ReservedRealmInteg
Fix CreateSnapshotRequestTests Failure (#31630)
Configurable password hashing algorithm/cost (#31234)
[TEST] Mute failing NamingConventionsTaskIT tests
[DOCS] Replace CONFIG_DIR with ES_PATH_CONF (#31635)
Core: Require all actions have a Task (#31627)
Support multiple system store types
When falling back to using the system keystore and - most usually -
truststore, do not assume that it will be a JKS store, but deduct
its type from {@code KeyStore#getDefaultKeyStoreType}. This allows
the use of any store type the Security Provider supports by setting
the keystore.type java security property.
As part of the changes in #31234,the password verification logic
determines the algorithm used for hashing the password from the
format of the stored password hash itself. Thus, it is generally
possible to validate a password even if it's associated stored hash
was not created with the same algorithm than the one currently set
in the settings.
At the same time, we introduced a check for incoming client change
password requests to make sure that the request's password is hashed
with the same algorithm that is configured to be used in the node
settings.
In the spirit of randomizing the algorithms used, the
{@code SecurityClient} used in the {@code NativeRealmIntegTests} and
{@code ReservedRealmIntegTests} would send all requests dealing with
user passwords by randomly selecting a hashing algorithm each time.
This meant that some change password requests were using a different
password hashing algorithm than the one used for the node and the
request would fail.
This commit changes this behavior in the two aforementioned Integ
tests to use the same password hashing algorithm for the node and the
clients, no matter what the request is.
Resolves#31670
Make password hashing algorithm/cost configurable for the
stored passwords of users for the realms that this applies
(native, reserved). Replaces predefined choice of bcrypt with
cost factor 10.
This also introduces PBKDF2 with configurable cost
(number of iterations) as an algorithm option for password hashing
both for storing passwords and for the user cache.
Password hash validation algorithm selection takes into
consideration the stored hash prefix and only a specific number
of algorithnm and cost factor options for brypt and pbkdf2 are
whitelisted and can be selected in the relevant setting.
Martijn van Groningen
Nhat Nguyen
Dimitris Athanasiou
Costin Leau
hanbj
Christoph Büscher
Sohaib Iftikhar
David Roberts
Vladimir Dolzhenko
Alpar Torok
Alexander Reelsen
Costin Leau
Simon Willnauer
Hendrik Muhs
Michael Basnight
Andrei Stefan
Ioannis Kakavas
Ioannis Kakavas
Julie Tibshirani
Tal Levy
Zachary Tong