Previously, the timestamp was accurate only down to seconds. For smaller intervals,
this would truncate the previous file(s). Now, for small intervals, the number of
microseconds is appended to the file name.
Added the possibility to disable timestamps in the memory profile file names and use
increments of 1 instead.
Updated the memory profiling integration tests.
Would previously happen if memory profiling was already started.
Now using os.File.Stat beforehand to check if a handler of the file
is already active and confirm the status of profiling. This required
changing the type of the reference fileCPU from io.Closer to *os.File.
Asserting the type would have worked as well.
- merge StopChanMemProf with StopMemoryProfiling
- remove fileMEM and stopMemProf from struct and constructors
- add separate mutex for memory profiling, ensure thread safety
- handle all significant errors
- log error if StopMemoryProfiling fails during CoreS Shutdown
- ignore errors if profiling inactive in Shutdown and deferred Stop
- move validations inside V1 functions
- return error if StartMemoryProfiling already started
- return error if StopMemoryProfiling already stopped or never started
- close profiling loop on error, not the cgr-engine
- StopMemoryProfiling closes channel and profiling loop writes final profile
- rename Path to DirPath for mandatory field error
- rename memprof_nrfiles flag to memprof_maxfiles
- increase default memprof_interval
- consider MaxFiles <= 0 as unlimited
- move memory profiling logic after starting services
- use CoreService Start/StopMemoryProfiling in main
- remove final memory profile block (created by deferred Stop)
- convert MemProfiling to method on CoreService and rename to profileMemory
- use Ticker for recurrent actions instead of Timer
- compute mem_final.prof full path in StartMemoryProfiling
- suffix profile files with current time instead of numbers
- update dispatcher methods after changes
- move MemoryPrf from utils to cores, rename to MemoryProfilingParams
- add logs for starting/stopping profiling
cgr-engine.go:
- use filepath.Join instead of path.Join
- handle *CoreService.StopCPUProfiling error inside deferred function
- same with the error from *os.File.Close()
cores/core.go:
- StartCPUProfile now returns an io.Closer (as opposed to an io.WriteCloser)
- handle pprof.StartCPUProfile error and ensure file is closed before returning
- log file close error as a warning if it occurs
- return missing mandatory error with correct path field name ('DirPath')
- no need to check if fileCPU is nil for profiling status
- pprof.StartCPUProfiling will return an error if profiling is already started
- os.File.Close() will return ErrClosed if profiling is already stopped
- differentiate between calling StopCPUProfiling when profiling hasn't started
and when it was already stopped by returning appropriate errors
- improved comments and error messages
- hardcode the base path for profiling endpoints to '/debug/pprof/'
- change profiling flag 'httprof_path' to boolean 'http_pprof'
(because of the above)
- remove redundant profiling endpoint registrations (handled by pprof.Index)
- move profiling registration log after actual registration
- make profiling registration log more descriptive
- use utils.Logger instead of log.Print for the log mentioned above
- refactor flags test into a table test (adding also verification for default
flag values)
- change 'scheduledShutdown' flag type from string to time.Duration (to avoid
unnecessary time parsing)
- revise flags usage descriptions
- rename flag 'singlecpu' to 'singleCPU'
- switch to 'ExitOnError' for flag parsing to simplify error handling and
automatically handle 'ErrHelp' by exiting with status 0 when help is
requested and status 2 for other parsing errors. Before the following error
log would have been received:
'<InitS> error received: <flag: help requested>, exiting!'
- update cgr-engine documentation
-update cached filter indexes count
-add the flaky constraint to a couple more tests
-fix unstable cores integration test
-fix unstable filterindexes integration test
Updated ees/ers implementation to use the jetstream package which
separates the jetstream context from Core NATS.
Removed the jsOpts fields from the NatsEE struct. We are now using the
jetStreamMaxWait option directly through a timeout context.
Added streamName option for NATS reader since it is now required to be
specified when creating a consumer (it is not inferred based on subject
anymore).
Updated nats ers integration test.
Removed deprecated birpc unit test.
Updated tests to also use the new jetstream package.
Updated tests to start the nats-server using their official driver
instead of using the std go exec package. time.Sleeps are now not
required anymore to wait for the server.
In test configurations for nats readers, made sure that natsStreamName
option is populated. It is now required for consumers to know where to
subscribe.
Fixed potential panic that happened when jetstreamMaxWait option would
not be set.
`ClientConnector` is no longer defined within `rpcclient` in its latest
version. It has been changed to be obtained from the `cgrates/birpc`
library instead.
Replaced `net/rpc` with `cgrates/birpc` and `net/rpc/jsonrpc` with
`cgrates/birpc/jsonrpc` libraries.
The implementations of `CallBiRPC()` and `Handlers()` were removed,
along with the methods associated with them.
The `rpcclient.BIRPCConector` and the methods prefixed with `BiRPC` were
removed from the `BiRPClient` interface.
The `BiRPClient` interface was renamed to `BIRPCClient`, although not
sure if needed (seems useful just to test if the structure is correct).
`rpcclient.BiRPCConector` has been replaced with `context.ClientConnector`,
which is now passed alongside `context.Context` within the same struct
(`cgrates/birpc/context.Context`). Consequently, functions that were
previously relying on it are now receiving the context instead. The
changes were made in the following functions:
- `engine/connmanager.go` - `*ConnManager.Call`
- `engine/connmanager.go` - `*ConnManager.getConn`
- `engine/connmanager.go` - `*ConnManager.getConnWithConfig`
- `engine/libengine.go` - `NewRPCPool`
- `engine/libengine.go` - `NewRPCConnection`
- `agents/libagents.go` - `processRequest`
Compilation errors related to the `rpcclient.NewRPCClient` function were
resolved by adding the missing `context`, `max_reconnect_interval`, and
`delayFunc` parameters. Additionally, context was added to all calls made
by the client. An effort was made to avoid passing hardcoded values as
much as possible, and extra flags were added where necessary for cgr
binaries.
The `max_reconnect_interval` parameter is now passed from parent
functions, which required adjustments to the function signature.
A new context field was added to all agent objects to ensure access to
it before sending it to the `connmanager's Call`, effectively replacing
`birpcclient`. Although an alternative would have been to create the
new service and add it to the context right before passing it to the
handlers, the chosen approach is definitely more comfortable.
With the addition of a context field for the SIP servers agents, an
additional error needed to be handled, coming from the creation of the
service. Agent constructors within the services package log errors as
they occur and return. Alternate solutions considered were either
shutting down the engine instead of returning, or just logging the
occurrence as a warning, particularly when the `ctx.Client` isn't
required, especially in cases where bidirectional connections are not
needed. For the latter option, it's crucial to return the object with
the error rather than nil or to make the error nil immediately after
logging.
Context has been integrated into all internal Call implementations to
ensure the objects conform to the `birpc.ClientConnector` interface.
These implementations will be removed in the near future as all service
objects are being wrapped in a `birpc.Service` type that satisfies the
`birpc.ClientConnector` interface. Currently, they are being retained
as a reference in case of any unexpected issues that arise.
Ensured that the `birpc.Service` wrapped service objects are passed to
the internal channel getters rather than the objects themselves.
Add context.TODO() to all \*ConnManager.Call function calls. To be
replaced with the context passed to the Method, when available.
For all `*ConnManager.Call` function calls, `context.TODO()` has been
added. This will be replaced with the context passed to the method when
it becomes available.
The value returned by StringGetOpts is now passed directly to the
FirstNonEmpty function, instead of being assigned to a variable
first.
The implementation of the `*AnalyzerService.GetInternalBiRPCCodec`
function has been removed from the services package. Additionally,
the AnalyzerBiRPCConnector type definition and its associated methods
have been removed.
The codec implementation has been revised to include the following
changes:
- `rpc.ServerCodec` -> `birpc.ServerCodec`;
- `rpc2.ServerCodec` -> `birpc.BirpcCodec`;
- `rpc2.Request` -> `birpc.Request`;
- `rpc2.Response` -> `birpc.Response`;
- The constructors for json and gob birpc codecs in `cenkalti/rpc`
have been replaced with ones from the `birpc/jsonrpc` library;
- The gob codec implementation has been removed in favor of the
version already implemented in the birpc external library.
The server implementation has been updated with the following changes:
- A field that represents a simple RPC server has been added to the
Server struct;
- Both the simple and bidirectional RPC servers are now initialized
inside the Server constructor, eliminating the need for nil checks;
- Usage of `net/rpc` and `cenkalti/rpc2` has been replaced with
`cgrates/birpc`;
- Additional `(Bi)RPCUnregisterName` methods have been added;
- The implementations for (bi)json/gob servers have been somewhat
simplified.
Before deleting the Call functions and using the `birpc.NewService`
method to register the methods for all cgrates components, update the
Call functions to satisfy the `birpc.ClientConnector` interface. This
way it will be a bit safer. Had to be done for SessionS though.
The `BiRPCCall` method has been removed from coreutils.go. The
`RPCCall` and `APIerRPCCall` methods are also to be removed in the
future.
Ensured that all methods for `SessionSv1` and `SessionS` have the
correct function signature with context. The same adjustments were made
for the session dispatcher methods and for the `SessionSv1Interface`.
Also removed sessionsbirpc.go and smgbirpc.go files.
Implemented the following methods to help with the registration of
methods across all subsystems:
- `NewServiceWithName`;
- `NewDispatcherService` for all dispatcher methods;
- `NewService` for the remaining methods that are already named
correctly.
Compared to the constructor from the external library, these also make
sure that the naming of the methods is consistent with our constants.
Added context to the Call methods for the mock client connectors (used
in tests).
Removed unused rpc fields from inside the following services:
- EeS
- LoaderS
- ResourceS
- RouteS
- StatS
- ThresholdS
- SessionS
- CoreS
Updated the methods implementing the logic for API methods to align
with the latest changes, ensuring consistency and correctness. The
modifications include:
- Adjusting the function signature to the new format
(ctx, args, reply).
- Prefixing names with 'V*' to indicate that they are utilized by
or registered as APIs.
- Containing the complete logic within the methods, enabling APIs
to call them and return their reply directly.
The subsystems affected by these changes are detailed as follows:
- CoreS: Additional methods were implementing utilizing the
existing ones. Though modifying them directly was possible, certain
methods (e.g., StopCPUProfiling()) were used elsewhere and not as
RPC requests.
- CDRs: Renamed V1CountCDRs to V1GetCDRsCount.
- StatS: V1GetQueueFloatMetrics, V1GetQueueStringMetrics,
V1GetStatQueue accept different arguments compared to API functions
(opted to register StatSv1 instead).
- ResourceS: Renamed V1ResourcesForEvent to V1GetResourcesForEvent
to align with API naming.
- DispatcherS: Renamed V1GetProfilesForEvent to
DispatcherSv1GetProfilesForEvent.
- For the rest, adding context to the function signature was enough.
In the unit tests, wrapping the object within a biprc.Service is now
ensured before passing it to the internal connections map under the
corresponding key.
Some tests that are covering error cases, are also checking the other
return value besides the error. That check has been removed since it
is redundant.
Revised the RPC/BiRPC clients' constructors (for use in tests)
A different approach has been chosen for the handling of ping methods
within subsystems. Instead of defining the same structure in every file,
the ping methods were added inside the Service constructor function.
Though the existing Ping methods were left as they were, they will be
removed in the future.
An additional method has been implemented to register the Ping method
from outside of the engine package.
Implemented Sleep and CapsError methods for SessionS (before they were
exclusively for bidirectional use, I believe).
A specific issue has been fixed within the CapsError SessionSv1 API
implementation, which is designed to overwrite methods that cannot be
allocated due to the threshold limit being reached. Previously, it was
deallocating when writing the response, even when a spot hadn't been
allocated in the first place (due to the cap being hit). The reason
behind this, especially why the test was passing before, still needs
to be looked into, as the problem should have occurred from before.
Implement `*SessionSv1.RegisterInternalBiJSONConn` method in apier.
All agent methods have been registered under the SessionSv1 name. For
the correct method names, the leading "V1" prefix has been trimmed
using the `birpc.NewServiceWithMethodsRename` function.
Revise the RegisterRpcParams function to populate the parameters
while relying on the `*birpc.Service` type instead. This will
automatically also deal with the validation. At the moment,
any error encountered is logged without being returned. Might
be changed in the future.
Inside the cgrRPCAction function, `mapstructure.Decode`'s output parameter
is now guaranteed to always be a pointer.
Updated go.mod and go.sum.
Fixed some typos.
