Make SSLContext reloadable
This commit replaces all customKeyManagers and TrustManagers
(ReloadableKeyManager,ReloadableTrustManager,
EmptyKeyManager, EmptyTrustManager) with instances of
X509ExtendedKeyManager and X509ExtendedTrustManager.
This change was triggered by the effort to allow Elasticsearch to
run in a FIPS-140 environment. In JVMs running in FIPS approved
mode, only SunJSSE TrustManagers and KeyManagers can be used.
Reloadability is now ensured by a volatile instance of SSLContext
in SSLContectHolder.
SSLConfigurationReloaderTests use the reloadable SSLContext to
initialize HTTP Clients and Servers and use these for testing the
key material and trust relations.
This commit is related to #28898. It adds an nio driven http server
transport. Currently it only supports basic http features. Cors,
pipeling, and read timeouts will need to be added in future PRs.
Due to the way composite aggregation works, ordering in GROUP BY can be
applied only through grouped columns which now the analyzer verifier
enforces.
Fix 29900
This commit removes the SecurityLifecycleService, relegating its former
functions of listening for cluster state updates to SecurityIndexManager
and IndexAuditTrail.
This change adds a grok_pattern field to the GET categories API
output in ML. It's calculated using the regex and examples in the
categorization result, and applying a list of candidate Grok
patterns to the bits in between the tokens that are considered to
define the category.
This can currently be considered a prototype, as the Grok patterns
it produces are not optimal. However, enough people have said it
would be useful for it to be worthwhile exposing it as experimental
functionality for interested parties to try out.
This is fixing an issue that has come up in some builds. In some
scenarios I see an assertion failure that we are trying to move to
application mode when we are not in handshake mode. What I think is
happening is that we are in handshake mode and have received the
completed handshake message AND an application message. While reading in
handshake mode we switch to application mode. However, there is still
data to be consumed so we attempt to continue to read in handshake mode.
This leads to us attempting to move to application mode again throwing
an assertion.
This commit fixes this by immediatly exiting the handshake mode read
method if we are not longer in handshake mode. Additionally if we swap
modes during a read we attempt to read with the new mode to see if there
is data that needs to be handled.
The errors were caused because release tests would use a copy of
the public key that was formatted differently. The change to the
public key format was introduced in [1].
Release tests Jenkins job has now been updated to use the correct
key format depending on the branch they run on [2]
Closes#30430
[1] https://github.com/elastic/elasticsearch/pull/30251
[2] https://github.com/elastic/infra/pull/4944
Since adding back the per-watch statistics, we do not need to access
every trigger engine implementation to get the current total job count.
This commit removes the unused methods to do so.
Tweak the return data, in particular with regards for ODBC columns to
better align with the spec
Fix order for SYS TYPES and TABLES according to the JDBC/ODBC spec
Fix#30386Fix#30521
This commit cleans up some code in the FileUserPasswdStore and the
FileUserRolesStore classes. The maps used in these classes are volatile
so we need to make sure that we don't perform multiple operations with
the map unless we are sure we are using a reference to the same map.
The maps are also never null, but there were a few null checks in the
code that were not needed. These checks have been removed.
The TokenMetaData equals method compared byte arrays using `.equals` on
the arrays themselves, which is the equivalent of an `==` check. This
means that a seperate byte[] with the same contents would not be
considered equivalent to the existing one, even though it should be.
The method has been updated to use `Array#equals` and similarly the
hashcode method has been updated to call `Arrays#hashCode` instead of
calling hashcode on the array itself.
These tests are both in the file `watcher/stats/10_basic`, and have been
failing fairly frequently over the last month with a start-up issue.
The issue is being tracked in #30298.
Dates internally contain milliseconds (which appear when converting them
to Strings) however parsing does not accept them (and is being strict).
The parser has been changed so that Date is mandatory but the time
(including its fractions such as millis) are optional.
Fix#30002
This commit adds a general state listener to the SecurityIndexManager,
and replaces the existing health and up-to-date listeners with that. It
also moves helper methods relating to health to SecurityIndexManager
from SecurityLifecycleService.
As conformance to best practices, this changes ensures that if a
SAML Response is signed, we verify the signature before processing
it any further. We were only checking the InResponseTo and
Destination attributes before potential signature validation but
there was no reason to do that up front either.
With the opening of xpack, we still retained a run task within
:x-pack:plugin. However, the root level run task also runs with the
default distribution. This change removes the extra run task inside
xpack in favor of using the root level task, and moves the
license/configuration code for run into the main run configuration.
This commit removes the hardcoded list of unconfigured ciphers in the
SslIntegrationTests. This list may include ciphers that are not
supported on certain JVMs. This list is replaced with code that
dynamically computes the set of ciphers that are not configured for
use by default.
The HTTPClient used in watcher is based on the apache http client. The
current client is using a lot of defaults - which are not always
optimal. Two of those defaults are the maximum number of total
connections and the maximum number of connections to a single route.
If one of those limits is reached, the HTTPClient waits for a connection
to be finished thus acting in a blocking fashion. In order to prevent
this when many requests are being executed, we increase the limit of
total connections as well as the connections per route (a route is
basically an endpoint, which also contains proxy information, not
containing an URL, just hosts).
On top of that an additional option has been set to evict
long running connections, which can potentially be reused after some
time. As this requires an additional background thread, this required
some changes to ensure that the httpclient is closed properly. Also the
timeout for this can be configured.
This commit renames IndexLifecycleManager to SecurityIndexManager as it
is not actually a general purpose class, but specific to security. It
also removes indirection in code calling the lifecycle service, instead
calling the security index manager directly.
Starting watcher should wait for the watcher to be started before
marking the status as started, which is now done via a callback.
Also, reloading watcher could set the execution service to paused. This could
lead to watches not being executed, when run in tests. This fix does not
change the paused flag in the execution service, just clears out the
current queue and executions.
Closes#30381
When validating the search request, we make sure any date_histogram
aggregations have timezones that match the jobs. But we didn't
do any such validation on range queries.
While it wouldn't produce incorrect results, it would be confusing
to the user as no documents would match the aggregation (because we
add a filter clause on the timezone for the agg).
Now the user gets an exception up front, and some helpful text about
why the range query didnt match, and which timezones are acceptable
Upgrade to lucene-7.4.0-snapshot-1ed95c097b
This version contains:
* An Analyzer for Korean
* An IntervalQuery and IntervalsSource that retrieve minimum intervals of positional queries.
* A new API to retrieve matches (offsets and positions) of a query for a single document.
* Support for soft deletes in the index writer.
* A fixed shingle filter that handles index time synonyms.
* Support for emoji sequence in ICUTokenizer (with an upgrade to icu 61.1)
When the watcher service pauses execution due to a cluster state update,
the trigger service and its engines also need to pause properly instead
of keeping going. This is also important when the .watches index is
deleted, so that watches don't stay in a triggered mode.
The IndexAndAliasesResolver resolves the indices and aliases for each
request and also handles local and remote indices. The current
implementation uses the ResolvedIndices class to hold the resolved
indices and aliases. While evaluating the indices and aliases against
the user's permissions, the final value for ResolvedIndices is
constructed. Prior to this change, this was done by creating a
ResolvedIndices for the first set of indices and for each additional
addition, a new ResolvedIndices object is created and merged with
the existing one. With a small number of indices and aliases this does
not pose a large problem; however as the number of indices/aliases
grows more list allocations and array copies are needed resulting in a
large amount of garbage and severely impacted performance.
This change introduces a builder for ResolvedIndices that appends to
mutable lists until the final value has been constructed, which will
ultimately reduce the amount of garbage generated by this code.
When dealing with filtering, a composite aggregation might return empty
buckets (which have been filtered) which gets sent as is to the client.
Unfortunately this interprets the response as no more data instead of
retrying.
This now has changed and the listener keeps retrying until either the
query has ended or data passes the filter.
Fix#30292
This commit fixes an issue with the data diagnostics were
empty buckets are not reported even though they should. Once
a job is reopened, the diagnostics do not get initialized from
the current data counts (especially the latest record timestamp).
The result is that if the data that is sent have a time gap compared
to the previous ones, that gap is not accounted for in the empty bucket
count.
This commit fixes that by initializing the diagnostics with the current
data counts.
Closes#30080
The current implementation starts/stops watcher using an executor. This
can result in our of order operations.
This commit reduces those executor calls to an absolute minimum in order
to be able to do state changes within the cluster state listener method,
which runs in sequence.
When a state change occurs that forces the watcher service to pause
(like no watcher index, no master node, no local shards), the service is
now in a paused state.
Pausing is a super lightweight operation, which marks the
ExecutionService as paused and waits for the currently executing watches
to finish in the background via an executor. The same applies for
stopping, the potentially long running operation is outsourced in to an
executor, as waiting for executed watches is decoupled from the current
state.
The only other long running operation is starting, where watches need to
be loaded. This is also done via an executor, but has an additional
protection by checking the cluster state version it was started with. If
another cluster state version was trying to load the watches, then this
loading will not take effect.
This PR also cleans up some unused states, like the a simple boolean in
the HistoryStore/TriggeredWatchStore marking it as started or stopped,
as this can now be caught in the execution service.
Another advantage of this approach is the fact, that now only triggered
watches are not getting executed, while watches that are run via the
Execute Watch API will still be executed regardless if watcher is
stopped or not.
Lastly the TickerScheduleTriggerEngine thread now only starts on data nodes.
This commit is a follow up to #30135. It updates the stream
compatibility versions in the start_trial requests and responses to
reflect that fact that this work has been backported to 6.3.
Necessary changes so that the licensing functionality can be
used in a JVM in FIPS 140 approved mode.
* Uses adequate salt length in encryption
* Changes key derivation to PBKDF2WithHmacSHA512 from a custom
approach with SHA512 and manual key stretching
* Removes redundant manual padding
Other relevant changes:
* Uses the SAH512 hash instead of the encrypted key bytes as the
key fingerprint to be included in the license specification
* Removes the explicit verification check of the encryption key
as this is implicitly checked in signature verification.
This commit removes the http.enabled setting. While all real nodes (started with bin/elasticsearch) will always have an http binding, there are many tests that rely on the quickness of not actually needing to bind to 2 ports. For this case, the MockHttpTransport.TestPlugin provides a dummy http transport implementation which is used by default in ESIntegTestCase.
closes#12792
* SQL: Reduce number of ranges generated for comparisons
Rewrote optimization rule for combining ranges by improving the
detection of binary comparisons in a tree to better combine
them in a range, regardless of their place inside an expression.
Additionally, improve the comparisons of Numbers of different types
Also, improve reassembly of conjunction/disjunction into balanced
trees.
Do not promote BinaryComparisons to Ranges since it introduces NULL
boundaries and thus a corner-case that needs too much handling
Compare BinaryComparisons directly between themselves and to Ranges
Fix#30017
The elasticsearch-users utility had various messages that were
outdated or incorrect. This commit updates the output from this
command to reflect current terminology and configuration.
The variadic constructor was only used in a few places and the
RepositoriesMetaData class is backed by a List anyway, so just using a
List will make it simpler to instantiate it.
Cause the CLI to ignore commands that are empty or consist only of
newlines. This is a fairly standard thing for SQL CLIs to do.
It looks like:
```
sql> ;
sql>
|
| ;
sql> exit;
Bye!
```
I think I *could* have implemented this with a `CliCommand` that throws
out empty string but it felt simpler to bake it in to the `CliRepl`.
Closes#30000
xpack core contains a fork of `Cron` from quartz who's javadoc has a
`<table>` with non-html5 compatible stuff. This html5ifies the table and
switches the `:x-pack:plugin:core` project to building javadoc with
HTML5.
This commit refactors the DataStreamDiagnostics class
achieving the following advantages:
- simpler code; by encapsulating the moving bucket histogram
into its own class
- better performance; by using an array to store the buckets
instead of a map
- explicit handling of gap buckets; in preparation of fixing #30080
Starting with the refactoring in https://github.com/elastic/elasticsearch/pull/22778 (released in 5.3) we may fail to properly replicate operation when a mapping update on master fails. If a bulk
operations needs a mapping update half way, it will send a request to the master before continuing
to index the operations. If that request times out or isn't acked (i.e., even one node in the cluster
didn't process it within 30s), we end up throwing the exception and aborting the entire bulk. This is
a problem because all operations that were processed so far are not replicated any more to the
replicas. Although these operations were never "acked" to the user (we threw an error) it cause the
local checkpoint on the replicas to lag (on 6.x) and the primary and replica to diverge.
This PR does a couple of things:
1) Most importantly, treat *any* mapping update failure as a document level failure, meaning only
the relevant indexing operation will fail.
2) Removes the mapping update callbacks from `IndexShard.applyIndexOperationOnPrimary` and
similar methods for simpler execution. We don't use exceptions any more when a mapping
update was successful.
I think we need to do more work here (the fact that a single slow node can prevent those mappings
updates from being acked and thus fail operations is bad), but I want to keep this as small as I can
(it is already too big).
Ioannis Kakavas
Tim Brooks
Costin Leau
Costin Leau
Ryan Ernst
Ioannis Kakavas
David Roberts
Igor Motov
David Kyle
Alexander Reelsen
Jay Modi
Julie Tibshirani
Nhat Nguyen
Jack Conradson
Zachary Tong
Jim Ferenczi
Dimitris Athanasiou
Tim Vernum
Michael Basnight
Nik Everett
Boaz Leskes