I see occasional confusion about the explanations emitted by the same-shard
allocation decider, particularly amongst new users setting up a single-node
cluster and trying to determine why their cluster has `yellow` health. For
example:
the shard cannot be allocated to the same node on which a copy of the shard
already exists
This is technically correct but it's quite a complicated sentence. Also, by
starting with "the shard cannot be allocated" it makes it sound like this is
the problem, whereas in fact this message is a good thing and users should
typically focus their attention elsewhere.
This commit simplifies the wording of these messages and makes them sound more
positive, for example:
a copy of this shard is already allocated to this node
In order to iterate through remote connections, the remote connection
manager maintains a local cache of connected nodes. Unfortunately this
is difficult in relationship with testing as it is inherently racy in
comparison to the parent connection manager map of connections.
This commit improves the relationship by only returning a cached
connection if it is still registered with the parent. If the connection
is not open, we will go to the slow path of allocating a iterator
directly from the parent.
* Emit deprecation warning if multiple v1 templates match with a new index (#55558)
* Emit deprecation warning if multiple v1 templates match with a new index
* DEPRECATION_LOGGER rename
This commit includes a number of minor improvements around `DelayableWriteable`: javadocs were expanded and reworded, `get` was renamed to `expand` and `DelayableWriteable` no longer implements `Supplier`. Also a couple of methods are now private instead of package private.
We generate a circular reference exception in translog in 6.8 in the
following scenario:
- The first rollGeneration hits "too many open files" exception when it's
copying a checkpoint file. We will set the tragic exception and close
the translog
- The second rollGeneration hits AlreadyClosedException as the current
writer is closed. We will suppress the ACE to the current tragic
exception. Unfortunately, this leads to a circular reference as ACE
already suppresses the tragic exception.
Other factors that help to manifest this bug:
- We do not fail the engine on AlreadyClosedException in flush
- We do not check for ensureOpen before rolling a new generation
Closes#55893
Ocassionally this test can fail when the randomized index.version.created is
before 6.1. In this case we don't check that if mappedFields.size() == 0 we
expect a MatchNoDocsQuery query being returned, which we do for other versions.
This fails only occasionally but with the seed provided on the original issue.
It also shouldn't be an issue on master since we shouldn't test with these pre-7
index versions there.
Closes#55950
Backports #55826 to 7.x
Modified AggregatorTestCase.searchAndReduce() method so that it returns an empty aggregation result when no documents have been inserted.
Also refactored several aggregation tests so they do not re-implement method AggregatorTestCase.testCase()
Fixes#55824
Currently if we shortcircuit a message the breaker release is null since
there is nothing to be broken. However, the TcpTransportChannel
infrastructure still expects it. This commit resolves this issue be
returning a no-op breaker release.
Currently `currentFieldType` is an instance variable that is first set and then
used by all methods referring to it. We can make it local to each method
instead, avoiding possible state problems and improve readability of the code
instead.
This adds a new api to simulate matching the given index name against the
index templates in the system.
The syntax for the new API takes the following form:
POST _index_template/_simulate_index/{index_name}
{
"index_patterns": ["logs-*"],
"priority": 15,
"template": {
"settings": {
"number_of_shards": 3
}
...
}
}
Where the body is optional, but we support the entire body used by the
PUT _index_template/{name} api. When the body is specified we'll simulate
matching the given index against a system that'd have the given index
template together with the index templates that exist in the system.
The response, in both cases, will return the matching template's resolved
settings, mappings and aliases, together with a special field that'll print any
overlapping templates and their corresponding index patterns.
(cherry picked from commit 1a5845edce1f445c58e094e9a3b6792e21e543b0)
Signed-off-by: Andrei Dan <andrei.dan@elastic.co>
Implements Sum aggregation over Histogram fields by summing the value of each bucket multiplied by their count as requested in #53285
Backports #55681 to 7.x
Today if state recovery is delayed by the `gateway.recover_after_*` settings
then we may end up performing state recovery twice: once when enough nodes have
joined the cluster, and again when the timeout elapses. The second state
recovery reinitializes the routing table, effectively discarding all
recovered/recovering shards and starting again from scratch. This commit adds a
check to prevent this second state recovery.
Closes#55564
Currently the testTransientErrorsDuringRecoveryAreRetried validates that
the expected peer recovery starts only once. This check is coarse and is
executed on all nodes and indexes. This commit modifies this check to
only be performed on the expected index. Additionally this commit
removes the disruption behavior from the "blue" node where it is not
relevant. Finally, this commit improves the logging for this test.
This replaces a reference to the result of partially reducing
aggregations that async search keeps with a reference to the serialized
form of the result of the partial reduction which we need to keep
anyway.
When calling scripts in metric aggregation, the returned metric state is
passed along to the coordinating node to do the final reduce. However,
it is possible the object could contain nested state which is unknown to
StreamOutput/StreamInput. This would then result in the node crashing as
exceptions are not expected in the middle of serialization.
This commit adds a method to StreamOutput that can determine if an
object is writeable by the stream. It uses the same logic
writeGenericValue, special casing each of the supported collection types
to recursively determine if each contained value is itself writeable.
relates #54708
Currently cancelling the RetryableAction does not stop one last run from
being executed. This commit makes a best effort attempt to cancel a
scheduled retry and guards future executions from the action already
being completed.
Currently a failed peer recovery action will fail an recovery. This
includes when the recovery fails due to potentially short lived
transient issues such as rejected exceptions or circuit breaking
errors.
This commit adds the concept of a retryable action. A retryable action
will be retryed in face of certain errors. The action will be retried
after an exponentially increasing backoff period. After defined time,
the action will timeout.
This commit only implements retries for responses that indicate the
target node has NOT executed the action.
If we advance the global checkpoint during commit and sync that
checkpoint after commit, then the assertions in the test won't hold
because the deletion policy did not see the latest global checkpoint
but only the value before committing.
Closes#55680
The disk decider had special handling for the single data node case,
allowing any allocation (skipping watermark checks) for such clusters.
This special handling can now be avoided via a setting.
If a node client (or rather its underlying node) is closed then
any executions on it will just quietly fail as happens in #55660
via closing the nodes on the test thread and asynchronously using
a node client.
Closes#55660
For 7.x, we already implemented the `?prefer_v2_templates` flag and made V2 templates opt-in, so we
can relax the error when updating V1 templates to just a warning. This will still be a hard error
for 8.0+
Relates to #53101
This has no practical impact on users since frozen indices are the only
throttled indices today. However this has an impact on upcoming features
that would use search throttling.
Filtering out throttled indices made sense a couple years ago, but as
we're now improving support for slow requests with `_async_search` and
exploring ways to reduce storage costs, this feature has most likely
become a trap, that we'd like to not have with upcoming features that
would use search throttling.
Relates #54058
Currently there is a clear mechanism to stub sending a request through
the transport. However, this is limited to testing exceptions on the
sender side. This commit reworks our transport related testing
infrastructure to allow stubbing request handling on the receiving side.
Fixes confusing error message when unknown value type is specified in a terms
aggregation. Adds support for parsing "numeric" and "number" value types.
Fixes#55727
We make sure to filter shard generations for indices that are missing
from the metadata when finalizing a partial snapshot (from concurrent index deletion)
but we failed to account for the case where we manually build a fake metadata instance
for snapshots without the global state.
Fixed this by handling missing indices by skipping, same way we do it for filtering the
shard generations.
Relates #50234
With this change, we will always return true for can_match requests on
idle search shards; otherwise, some shards will never get refreshed if
all search requests perform the can_match phase (i.e., total shards >
pre_filter_shard_size).
Relates #27500
Relates #50043
David Turner
Mark Tozzi
Tim Brooks
Igor Motov
Lee Hinman
Przemko Robakowski
Luca Cavanna
Andrei Dan
Nhat Nguyen
Christoph Büscher
Dan Hermann
Christos Soulios
Mark Vieira
Yang Cheng
Armin Braun
Nik Everett
Ryan Ernst
Henning Andersen
Adrien Grand
weizijun
Julie Tibshirani
Tal Levy
Jim Ferenczi