The current delay waits until later after normal configuration, but
still just after resolution (ie when paths would be known for
dependencies but not actual execution). This delays the checks further
to be done right before we actually execute the copy task.
closes#15068
Currently if running with --info, the command line for ES, along with
env vars, are logged before they may be ammended to add debug options.
This moves the adding JAVA_OPTS to before we print the command.
Removed check that two query builder that are different according
to equals() have different hashCode since that is not required
by the contract of hashCode.
When importing dangling indices on a single node that is data and master eligable the async dangling index
call can still be in-flight when the cluster is checked for green / yellow. Adding a dedicated master node
and a data only node that does the importing fixes this issus just like we do in OldIndexBackwardsCompatibilityIT
We used to check on several places if we are still open but non of these
places did the check under the lock which leaves a small window where we
potentially get closed but still access an already closed channel or another
IO resource.
This is a pretty trivial change that moves most of the monitor service related
object creation from guice into the monitor service. This is a babystep towards removing
guice on the node level as well. Instead of opening huge PRs I try to do this in baby-steps
that are easier to digest.
The Exec task outputs stdout/stderr to the standard streams by default.
However, to keep output short, we currently capture this, and only
output if the task failed. This change makes a small wrapper around Exec
to facilitate this behavior anywhere we use Exec.
This commit adds a property that will prevent the Gradle daemon from
being used for builds (even if one is running). This is to avoid some
nasty issues (e.g., SIGBUS faults and other mmap diasters) that result
from class loaders not being closed properly.
We handle AlreadyClosedExceptions gracefully wherever IndexShard / Engine
is used. In some cases, instead of throwing the appropriate exception we
bubble up ChannelClosedException instead which causes shard failures etc.
Today, it seems like that this can only happen if the engine is closed without
acquireing the lock which means that the shard has failed already so the impact is really
just a confusing log message. Yet, this change enforces throwing the right exception
if the translog is already closed.
Closes#14866
Michael McCandless
Ryan Ernst
Nik Everett
Johnny Lim
Clinton Gormley
Jason Spafford
Robert Muir
Christoph Büscher
Daniel Mitterdorfer
Simon Willnauer
Robert Muir
Jason Tedor
Jayson Minard
Simon Willnauer
Britta Weber
tomoya yokota
Lee Hinman
Jimmi Dyson
Adrien Grand
Isabel Drost-Fromm