| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|\
| |
| | |
Filter on file creation time instead of last modified time
|
| | |
|
|/ |
|
|\
| |
| | |
Remove guice binding for Linguistics.
|
| | |
|
| |
| |
| |
| |
| | |
- The binding has been removed from HandlersConfigurerDi to
prevent creating a SimpleLinguistics there.
|
| |
| |
| |
| |
| | |
- The config model adds a SimpleLinguisticsProvider for all
container clusters.
|
| | |
|
| | |
|
|/ |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
vespa-engine/revert-6719-bratseth/initially-down"
This reverts commit 05ea4c2b9952bceaf0e009053c2f51a09d15a957, reversing
changes made to 03efdabdc9f0e43fc2c362569ae455a7abc9a64b.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Before ths change containers used to come up in the initial VIP state of up
("in rotation") before having any information about downstream cløusters.
This is problematic if the node determines that it should be
down when learning the state of downstream clusters, especially
when it is determining that it should be down due to failing to complete
pings timely due to initialization load.
This changes to starting in the down state. Crucially, this requires
tracking the current state of clusters in a separate component
such that the state survives reconfiguration and we avoid setting the
state down briefly on each reconfiguration.
In addition to the case where there are downstream clusters deciding
the VIP state oif a container, there are two others:
- There are no such clusters. In this case the VIP state will be up by default.
- Some other component is explicitly deciding the VIP state. This commit makes
that case clearer.
|
| |
|
|
|
|
|
|
|
|
| |
.. to prevent embedding them in container-disc and potentially
other bundles.
- Must be compile scope in linguistics because they're needed to
run tests in modules depending on linguistics, e.g.
indexinglanguage.
|
|\
| |
| | |
Tenant pipelines API
|
| | |
|
|\ \
| |/
|/| |
Bratseth/java model inference
|
| | |
|
| | |
|
| | |
|
| | |
|
|\ \ |
|
| | | |
|
| | | |
|
| |\ \
| | |/
| |/| |
Henrhoi/json query api
|
| | | |
|
|/ /
| |
| |
| | |
building queries. Resources-folder static is now named gui
|
| | |
|
|/ |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Found this edge-case with logAndDie in ThreadPoolProvider:
- A Runnable executes and finishes at time T0
- No Runnables execute for a long time t > maxThreadExecutionTimeSeconds
- At time T1 = T0 + t, a bursts of Runnables N > maxthreads arrives such that
the (maxthreads+1) Runnable will be tried executed and rejected before any
of the first maxthreads Runnables finish.
- Then, logOrDie will be called immediately.
That is, a sufficiently silent period followed by a burst of Runnables cause
the process to die.
This PR will make sure that the process dies only if
maxThreadExecutionTimeSeconds seconds pass with no available threads, like the
documentation says.
I found this as I was looking for reasons why the Cluster Controller have had
logAndDie problems. The CC has maxthreads of 10, and was getting bursts of 6-8
of a particular request type (reduced in PR 6207).
|
| |
|
|
|
|
|
| |
Use config value for inital value of isInRotation() when nothing
is known about backend clusters
|
| |
|
|
|
|
|
|
|
|
|
| |
Separate system internal redeploys from application package
changes such that we can ignore config updates originating
from application package changes if restart on deploy is true,
as the config change will be followed by a restart - but continue
to effect config changes originating from internal system changes
as those will not (and should not) be followed by a restart.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
* Ensure no empty chains before throwing IndexOutOfBoundsException.
* Verify that application does not throw upon access-control
in services.xml.
|
| |
|
|
|
|
|
|
|
|
| |
- The SizeCalculator uses recursive reflective calls in the object
tree, and attempts illegal access to e.g. java.lang.module
classes on Java 9+. With this access explicitly allowed via
"--add-opens" the calculator returns a much larger size than
the actual size of the object.
|
|\
| |
| | |
Gjoranv/java9 prep 06
|
| |
| |
| |
| |
| |
| |
| | |
- javax.xml classes are moved to module java.xml.ws
("which is not in the module graph")
- The cert-k test bundle referenced a lot of packages that are
moved to other modules, but are unlikely to be used.
|