Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Remove metrics for persistence provider that were deprecated and broken (not ↵ | Geir Storli | 2019-01-21 | 2 | -5/+1 |
| | | | | thread safe) on Vespa 6. | ||||
* | Remove workarounds added during transition to multiple bucket spaces. | Tor Egge | 2019-01-21 | 4 | -51/+9 |
| | |||||
* | Remove deprecated iterators_per_bucket setting from stor-visitor.def. | Tor Egge | 2019-01-21 | 2 | -50/+2 |
| | |||||
* | Remove deprecated memorytouse setting in stor-server.def. | Tor Egge | 2019-01-21 | 1 | -5/+0 |
| | |||||
* | Merge pull request #8153 from vespa-engine/havardpe/remove-fastlib-webserver | Henning Baldersheim | 2019-01-15 | 1 | -1/+0 |
|\ | | | | | remove fastlib webserver | ||||
| * | remove fastlib webserver | Håvard Pettersen | 2019-01-15 | 1 | -1/+0 |
| | | |||||
* | | Merge pull request #8109 from vespa-engine/vekterli/fix-tsan-detected-data-races | Tor Brede Vekterli | 2019-01-15 | 7 | -24/+30 |
|\ \ | |/ |/| | Fix some ThreadSanitizer reported data races | ||||
| * | Use atomics for flags and configs set and read across threads | Tor Brede Vekterli | 2019-01-10 | 5 | -16/+19 |
| | | |||||
| * | Use atomics to ensure well-defined memory visibility in test | Tor Brede Vekterli | 2019-01-10 | 1 | -7/+8 |
| | | |||||
| * | Pending count must be atomic, read and written from different threads | Tor Brede Vekterli | 2019-01-09 | 1 | -1/+3 |
| | | |||||
* | | improve path/query handling | Håvard Pettersen | 2019-01-14 | 1 | -14/+2 |
|/ | | | | | | - return attributes by value to avoid referencing deconstructed objects. - use path/query form portal request to avoid dequoting issues - log raw uri instead of dequoted non-parsed uri | ||||
* | GC unused code and update includes | Henning Baldersheim | 2019-01-05 | 2 | -1/+2 |
| | |||||
* | Quck bail out when port can not be listened to. | Henning Baldersheim | 2019-01-04 | 1 | -2/+8 |
| | |||||
* | log status server exceptions | Håvard Pettersen | 2019-01-04 | 1 | -0/+1 |
| | |||||
* | move-construct function member | Håvard Pettersen | 2019-01-04 | 1 | -1/+1 |
| | |||||
* | do not inherit from Runnable | Håvard Pettersen | 2019-01-04 | 1 | -3/+1 |
| | |||||
* | use portal in storage webserver | Håvard Pettersen | 2019-01-02 | 3 | -243/+114 |
| | |||||
* | Add TLS statistics to vespalib and expose as metrics via storageserver | Tor Brede Vekterli | 2018-12-18 | 6 | -8/+123 |
| | | | | Now without unused expiry time extraction. | ||||
* | Revert "Add TLS statistics to vespalib and expose as metrics via storageserver" | Harald Musum | 2018-12-18 | 6 | -123/+8 |
| | |||||
* | Add TLS statistics to vespalib and expose as metrics via storageserver | Tor Brede Vekterli | 2018-12-17 | 6 | -8/+123 |
| | | | | | Also add functionality for extracting "notAfter" expiration time from current certificate, which may later be added as an expiry metric. | ||||
* | Support test-and-set for auto-create document updates | Tor Brede Vekterli | 2018-11-26 | 7 | -29/+72 |
| | | | | | | | Has the obvious consistency caveats that if all your existing replicas are down, the update will go through since the document from an weak consistency perspective does not exist anywhere. But can be a useful feature if this is an acceptable tradeoff. | ||||
* | Remove experimental enable-multiple-bucket-spaces flag. | Geir Storli | 2018-11-23 | 6 | -50/+4 |
| | | | | The feature has been default on since late May 2018. | ||||
* | Correct copyright year | Tor Brede Vekterli | 2018-11-20 | 1 | -1/+1 |
| | |||||
* | Add explicit metric for diverging timestamps on updated documents | Tor Brede Vekterli | 2018-11-19 | 11 | -13/+95 |
| | | | | Easier to track across nodes than hunting for log warnings. | ||||
* | Warn on updates done against divergent timestamps | Tor Brede Vekterli | 2018-11-14 | 1 | -20/+20 |
| | | | | | | | | | | This should not happen during normal operation as the distributor should be aware if bucket replicas are out of sync before deciding whether an update requires a two-phase fallback. Warning for now to see if it can be observed anywhere. The logic surrounding the warning will likely be rewritten to treat this as an explicit failure at a later point. | ||||
* | Expose exception information. If totally unknown, generate core. | Henning Baldersheim | 2018-10-09 | 3 | -13/+1 |
| | |||||
* | remove legacy tags, always pass vector | Arne Juul | 2018-10-08 | 17 | -171/+191 |
| | |||||
* | Merge pull request #7193 from vespa-engine/arnej/drop-dimension-key | Arne H Juul | 2018-10-06 | 3 | -6/+6 |
|\ | | | | | Arnej/drop dimension key | ||||
| * | avoid using old "dimensionkey" concept | Arne Juul | 2018-10-03 | 3 | -6/+6 |
| | | | | | | | | | | | | | | | | | | | | * adding a "dimensionkey" would add an extra dimension to json reporting, but used the Metric name as the value in that dimension. This means there was no added information at all. Also, the "operationtype" dimension wasn't populated with operation types, even if it looked that way, because it wasn't used directly but was cloned and the actual name would be the loadtype instead, making it very confusing. | ||||
* | | Let CreateIterator inherit priority of parent visitor | Tor Brede Vekterli | 2018-10-05 | 2 | -2/+6 |
| | | | | | | | | | | Since CreateIterator now does more than just in-memory metadata work (i.e. it may hit the disk), it should not be given a fixed, very high priority. | ||||
* | | Remove the costly and unnecessary setSlobrokId interface. | Henning Baldersheim | 2018-10-04 | 1 | -15/+12 |
|/ | |||||
* | avoid "clever" operator++ overload | Arne Juul | 2018-09-26 | 8 | -38/+38 |
| | |||||
* | Remove BatchPutRemove command from storageapi. | Tor Egge | 2018-09-26 | 12 | -27/+5 |
| | |||||
* | Remove BatchDocumentUpdateMessage and BatchDocumentUpdateReply from | Tor Egge | 2018-09-25 | 6 | -69/+1 |
| | | | | | | C++ portion of documentapi. Remove corresponding document api cross language files. Remove corresponding storageapi messages. | ||||
* | remove non-instant invocation | Håvard Pettersen | 2018-09-05 | 1 | -5/+5 |
| | |||||
* | Remove unused code | Henning Baldersheim | 2018-08-20 | 4 | -33/+0 |
| | |||||
* | Deinline for gcc 8 | Henning Baldersheim | 2018-08-15 | 2 | -0/+3 |
| | |||||
* | Remove whitespace | Henning Baldersheim | 2018-08-12 | 7 | -10/+10 |
| | |||||
* | Pass stringref by value | Henning Baldersheim | 2018-08-11 | 3 | -3/+3 |
| | |||||
* | Pass stringref by value | Henning Baldersheim | 2018-08-10 | 7 | -10/+10 |
| | |||||
* | Deactivate all buckets in content layer bucket db when derived cluster | Tor Egge | 2018-08-07 | 4 | -12/+30 |
| | | | | | state indicates that node changes from up to down state even if node remains up according to baseline cluster state. | ||||
* | Remove stringref::c_str() | Tor Brede Vekterli | 2018-07-25 | 13 | -39/+40 |
| | | | | | | | | | | | | | The expected semantics of c_str() (a null-terminated string) cannot be satisfied with a string reference, so remove the function entirely to prevent people from using it in buggy ways. Replaces c_str() with data() in places where it is presumed safe, otherwise constructs temporary string instances. Certain callsites have been de-stringref'd in favor of regular strings, in particular where C APIs have been transitively called. The vast majority of these were called with string parameters anyway, so should not cause much extra allocation. | ||||
* | Test operation batch barrier with different lock modes | Tor Brede Vekterli | 2018-07-16 | 2 | -1/+20 |
| | |||||
* | Add tests of shared/exclusive persistence queue locking | Tor Brede Vekterli | 2018-07-16 | 5 | -74/+156 |
| | |||||
* | Use shared locking for Get/CreateIterator/GetIter operations | Tor Brede Vekterli | 2018-07-12 | 1 | -0/+6 |
| | |||||
* | Add support for exclusive/shared persistence bucket locking | Tor Brede Vekterli | 2018-07-12 | 5 | -56/+143 |
| | | | | | | | | | | | | | Operations that are tagged as only requiring shared locking may run concurrently with other shared lock operations. Operations requiring exclusive locking (all mutations) can't run concurrently with any other operations, shared or not. Let operations be able to explicitly specify their locking requirements. Default is exclusive locking, which shall give the same behavior as before. Since this locking happens at a stripe granularity, it's vital that operations are routed deterministically to the correct stripe based on their bucket. | ||||
* | Hardcode visitor iterators per bucket to 1 | Tor Brede Vekterli | 2018-07-12 | 3 | -30/+19 |
| | | | | | | | | | | We have an SPI iterate() invariant that states the provider shall never observe concurrent calls for the same iterator ID. If GetIter operations can operate in shared locking mode, this invariant will no longer hold if multiple GetIters for a single iterator ID can be pipelined in the queue. We therefore ignore the config entirely. This is not expected to cause any performance regressions in practice. | ||||
* | Merge pull request #6180 from vespa-engine/arnej/use-log-abort | Henning Baldersheim | 2018-06-21 | 12 | -17/+31 |
|\ | | | | | use LOG_ABORT not just abort() | ||||
| * | some more cleanup | Arne Juul | 2018-06-20 | 1 | -1/+0 |
| | | |||||
| * | move LOG_SETUP after other includes | Arne Juul | 2018-06-20 | 4 | -7/+11 |
| | |