| Commit message (Collapse) | Author | Age | Files | Lines |
| ... | |
| |
|
|
| |
feature: added oplog-mirror command. this can be invoked to export oplog contents to corresponding files
|
| |
|
| |
* Make sure we close our trace session properly
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* add update endpoint for project store project
* add update endpoint for oplog
* changelog
* Zen command line tool `project-update`
Zen command line tool `project-delete`
Zen command line tool `oplog-update`
Zen command line tool `oplog-delete`
* add --force-update option to project/oplog create
remove project/oplog update commnad
|
| |
|
|
|
| |
* Make sure to check oplog op attachments when gathering references for GC
* Add oplog op content to error result if attachment is missing when doing `oplog-export`
|
| | |
|
| |
|
|
| |
* rework cache store background jogging
* correct capture for context
|
| |
|
|
|
| |
* Skip upstream logic early if we have not upstream endpoints
* make cache store logging of CbObjects async
* changelog
|
| |
|
|
|
|
|
| |
* add basic stats for project store
* stats for BadRequest in ProjectStore::Rpc
* changelog
* group stats
fix ChunkWriteCount when accepting ops
|
| |
|
|
|
|
| |
* bump vcpkg version
* fmt lib 10 fixes
* xmake dependencies (with linux workarounds)
* changelog
|
| |
|
|
|
|
|
|
|
| |
entries (#355)
* Make sure we always write "data" attachment hash for snapshotted oplog entries
* Make sure to add chunk mappings for files moved to attatchment in snapshot operation
* fix inverted timoute for expiration (we don't want time expiry in these cases)
* increase timeout for jupiter oplog in project to 3 min
* changelog
|
| |
|
|
|
|
| |
* Add `response.text` to output in log when jupiter request fails
* only hash jupiter ref once and keep that
* Increase request timeout when uploading to Jupiter to 2 min (to handle very large attachments)
* changelog
|
| |
|
|
| |
specified (#352)
|
| | |
|
| | |
|
| |
|
|
| |
HTTP/2 endpoint without without HTTP/1.1 upgrade (#347)
|
| |
|
|
| |
* make sure we return an error code instead of throwing exception if header parsing fails
* changelog
|
| |
|
|
| |
* use streaming read for PutCompressedBlob if source is single file
* changelog
|
| |
|
|
| |
correctly to zenserver (#346)
|
| |
|
|
| |
* add more details in log if jupiter operations fail
* changelog
|
| |
|
|
|
|
| |
* Send proper error to caller of GetChunkInfo instead of assert
* catch and handle exceptions when checking for state_marker
* properly wait for background tasks if oplop-export fails
* changelog
|
| |
|
| |
* try to resolve access token via env variable local first, on zenserver side second
|
| |
|
| |
* make sure we validate pointers returned from zen::Memory::Alloc
|
| |
|
| |
* add more context for oplog import/export errors
|
| |
|
| |
* catch exceptions when scheduling GC and when writing GC scheduling state
|
| | |
|
| |
|
|
| |
also improved help message
|
| |
|
|
|
| |
this allows client code to use the ChunkResolver interface instead of
CidStore, which can help with testing scenarios
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
* Added LoadCompactBinaryObject from file to compactbinaryfile.cpp/h
* Added SaveCompactBinary(BinaryWriter& Ar, ...) functions
* Added ZEN_PLATFORM_NAME define
* Added SystemMetrics functionality to query system properties (see zencore/system.h)
|
| |
|
|
|
|
|
|
|
| |
* Made HttpHealthService use locks to serialize access to state
* Added ToString(HttpResponseCode HttpCode)
* Added support for JS source maps
* Moved IHttpStatsProvider/IHttpStatsService
* Enabled enumeration of stats providers
* Disabled build of HttpTestingService unless ZEN_WITH_TESTS is defined
|
| |
|
|
| |
cherry-picked from sb/proto to reduce delta
|
| |
|
|
|
| |
* added delete-on-close support to BasicFile
* added BasicFile::ReadRange()
|
| |
|
|
| |
IsProcessRunning on Windows would only consider if we could get a handle to a process. It is possible to get a handle to a process even if it is terminated in Windows. To actually know if the process is running, a further call to GetExitCodeProces is required. Addressing this issue ensures that the ZenServerState::Sweep method doesn't keep terminated processes in the state table.
|
| |
|
|
| |
newly spawned server instance
|
| |
|
|
| |
we don't currently permit creating zero-length IoBuffer instances to prevent inefficiencies, but this can cause issues since some higher level code now ends up using it for things it was not originally intended for.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
this change adds a serve command to the zen CLI. This can be used to establish links to a set of files which may be served to clients via the project store interface:
```cmd> zen serve Lyra/WindowsClient d:\temp_share\StagedBuilds\WindowsClient```
with the appropriate changes in UE you may then start an instance of the runtime and have it load all files via the remote file connection:
```
Lyra\Binaries\LyraClient.exe ../../../Lyra/Lyra.uproject -pak -basedir=D:\temp_share\StagedBuilds\WindowsClient/Lyra/Binaries/Win64 -Mount=Lyra/WindowsClient
```
|
| |
|
|
|
| |
this change adds `ZenServerInstance::SpawnServerAndWait()` which as the name implies spawns a server and then waits for the instance to reach a usable state before returning to the caller.
It also changes the behaviour of `ZenServerInstance::AttachToRunningServer()` so it matches the Spawn behaviour wrt automatic termination on ZenServerInstance destruction. Previously it would always terminate the subprocess on exit.
|
| | |
|
| |
|
|
|
| |
- curl is a dependency of zenhttp, not zencore
- zencore only depends on crypto functions from openssl
|
| |
|
|
| |
this enables mapping the result of a CreateProc call to an integer pid
|
| |
|
|
|
|
| |
* Improve resonse messages from jupiter upstream
changelog
* simplify response parsing in jupiter upstream
* changelog
|
| |
|
|
| |
characters (#332)
|
| |
|
|
| |
* add zen oplog-snapshot command line tool
* changelog
|
| |
|
|
| |
* make sure to prefix auth token correctly when provided with a raw token
* changelog
|
| | |
|
| |
|
|
| |
addresses ZEN-SERVER-W5. This should not be logged as an error since the failure is propagated to the client and the failure is typically due to invalid input
|