summaryrefslogtreecommitdiffstats
path: root/config/src/main/java/com
Commit message (Collapse)AuthorAgeFilesLines
...
* | Merge pull request #19280 from vespa-engine/bratseth/cleanup-876Harald Musum2021-09-272-3/+6
|\ \ | |/ |/| No functional changes
| * No functional changesJon Bratseth2021-09-242-3/+6
| |
* | Always send payload in responseHarald Musum2021-09-231-20/+8
|/ | | | | | | As an optimization we sent empty payload if config was unchanged. This optimization is no longer needed, as configs are small and we can remove checks several places in config system and ease changes to config protocol.
* Test with xxhash64 as wellHarald Musum2021-09-161-28/+19
|
* More cleanup of md5 usageHarald Musum2021-09-154-21/+4
|
* Remove getConfigMd5() from RawConfig, update usageHarald Musum2021-09-151-3/+0
|
* Revert "Revert "Update use of md5 and xxhash64 in more places [run-systemtest]""Harald Musum2021-09-154-8/+33
|
* Revert "Update use of md5 and xxhash64 in more places [run-systemtest]"Harald Musum2021-09-154-33/+8
|
* Update use of md5 and xxhash64 in more placesHarald Musum2021-09-144-8/+33
| | | | | | Set md5 and xxhash64 only when non-null in request and response. Update hasEqualConfig methods. Always return xxhash64 in response.
* Fix error in locig in previous commitHarald Musum2021-09-131-1/+1
| | | | Return early only if config has changed
* Remove md5 sum of config from initial state in config subscriptionHarald Musum2021-09-135-8/+40
| | | | | | Java applications will now request config with only xxhash64 of config, C++ applications will request config with md5 of config and config proxy will use what is in the request from the clients.
* Add checksum types in response based on types in requestHarald Musum2021-09-104-3/+45
| | | | | | | Return response with same type as in request, except when both are empty, then return both types. Note: No clients have started sending requests with only xxhash64 checksum yet.
* Stop validating length of checksumsHarald Musum2021-09-093-19/+10
| | | | | | | Not needed, and the way we calculated hex string from a long did not always return 16 characters, so fixes that issue. Also log validation failures as warnings
* Revert "Revert "Use both xxhash64 and md5 for config payload in requests and ↵Harald Musum2021-09-0921-97/+316
| | | | responses [run-systemtest]""
* Revert "Use both xxhash64 and md5 for config payload in requests and ↵Harald Musum2021-09-0921-316/+97
| | | | responses [run-systemtest]"
* Fix mistake made in previous commitHarald Musum2021-09-091-1/+1
|
* Use both xxhash64 and md5 for config payload in requests and responsesHarald Musum2021-09-0921-97/+316
| | | | Chacking of equal config with hasEqualConfig() still uses only md5
* Remove def md5 from ConfigKeyHarald Musum2021-09-0810-41/+65
|
* Update hex pattern to only contain hex charactersHarald Musum2021-09-071-1/+1
|
* GC unused code for setting lz4 compression levelHarald Musum2021-09-071-10/+4
|
* Update javadocHarald Musum2021-09-071-1/+1
|
* Extract method for validating checksumHarald Musum2021-09-073-21/+37
|
* Add class for payload checksumHarald Musum2021-09-067-31/+60
|
* Reforman, no functional changesHarald Musum2021-09-065-42/+57
|
* Deprecate and stop using ConfigDefinition constructor with versionHarald Musum2021-09-025-35/+36
|
* Refactor creation of config requestHarald Musum2021-06-211-9/+13
|
* Use constructor for command line argumentsHarald Musum2021-06-211-9/+27
|
* Rename methodHarald Musum2021-06-211-4/+5
|
* Validate configs before running load tester threadsHarald Musum2021-06-211-11/+16
|
* Rename variableHarald Musum2021-06-181-4/+4
|
* Use nanoTime()Harald Musum2021-06-181-9/+7
|
* Remove some debugging and measuring bytes/sec (only done when debugging)Harald Musum2021-06-181-14/+3
|
* Don't print a line for every requestHarald Musum2021-06-171-1/+0
|
* Revert "Revert "Cleanup, no functional changes, take 2""Harald Musum2021-06-171-38/+40
|
* Revert "Cleanup, no functional changes, take 2"Harald Musum2021-06-171-40/+38
|
* Cleanup, no functional changes, take 2Harald Musum2021-06-171-38/+40
| | | | Trying again, with fewer changes
* Revert "Cleanup, no functional changes"Harald Musum2021-06-161-54/+50
|
* Cleanup, no functional changesHarald Musum2021-06-161-50/+54
|
* Update JRTConnectionPool.javaHenning Baldersheim2021-06-151-1/+1
| | | Remove extraneous .
* Merge branch 'master' into hmusum/add-constructor-for-naming-JRTConnectionPoolHenning Baldersheim2021-06-151-1/+1
|\
| * drop empty buffers instead of using small buffersHåvard Pettersen2021-06-151-1/+1
| |
* | Add constructor for naming JRTConnectionPoolHarald Musum2021-06-151-3/+9
|/
* Do an explicit copy.Henning Baldersheim2021-05-081-1/+4
|
* Use AbstractUtf8Array in interface and a more flexible implementation to ↵Henning Baldersheim2021-05-075-23/+35
| | | | reduce amount of copying.
* Add ByteBuffer as interface.Henning Baldersheim2021-05-071-3/+10
|
* Add ByteBuffer to the interface to prepare for less copying of data.Henning Baldersheim2021-05-071-0/+10
|
* Remove redundant checks for closed subscriptionHarald Musum2021-04-301-6/+7
|
* Change delay when there are transient config errorsHarald Musum2021-04-302-20/+23
| | | | | | Delay next config request a random amount of time between 0 and error delay. This is to avoid thundering herd issues. Delay when getting fatal errors are unchanged.
* Simplify calculating delay after config failuresHarald Musum2021-04-292-46/+32
| | | | | | | | When there are several subscribers to a config and 1 JRTConfiguRequester the counter for failures might quickly increase so backoff based on these numbers will not be useful. Simplify by using booleans for failures instead. Max delay for configured subscriptions will now be 30 seconds instead of 150 seconds (which is way too much)
* More lazy debug log message generationJon Marius Venstad2021-04-281-3/+3
|