PDiskLogLoad

All VDisks hosted on a certain PDisk log data about their own performance to the common PDisk log. VDisks gradually delete their obsolete data at the beginning of the log to free up disk space. Sometimes, after one VDisk completes logging data and before another one starts logging it, a section with useless obsolete data may appear. In this case, such data is deleted automatically, and the PDiskLogLoad actor will perform a test to check whether such an operation is running correctly.

SizeInterval
VDisk1
SizeInterv...
BurstSize VDisk1
BurstSize VDisk1
BurstInterval
VDisk1
BurstInter...
StorageDuration VDisk1
StorageDuration VDisk1
SizeInterval
VDisk2
SizeInterv...
BurstInterval
VDisk2
BurstInter...
StorageDuration VDisk2
StorageDuration VDisk2
BurstSize VDisk2
BurstSize VDisk2
Data that is being deleted
Data that is being dele...
Timeline,
bytes
Timeline,...
Text is not SVG - cannot display

Note

This ad-hoc actor is used for testing specific functionality. This is not a load actor. It is designed to check whether something works properly.

Actor parameters

The basic actor parameters are described below. For the full list of parameters, see the load_test.proto file in the YDB Git repository.

Parameter Description
PDiskId ID of the Pdisk being loaded on the node.
PDiskGuid Globally unique ID of the PDisk being loaded.
VDiskId Parameters of the VDisk used to generate load.
  • GroupID: Group ID.
  • GroupGeneration: Group generation.
  • Ring: Group ring ID.
  • Domain: Ring fail domain ID.
  • VDisk: Index of the VDisk in the fail domain.
MaxInFlight Number of simultaneously processed requests.
SizeIntervalMin Minimum size of log record in bytes.
SizeIntervalMax Maximum size of log record in bytes.
BurstInterval Interval between logging sessions in bytes.
BurstSize Total amount of data to log per session, in bytes.
StorageDuration Virtual time in bytes. Indicates how long the VDisk should store its data in the log.
IsWardenlessTest Set it to False in case the PDiskReadLoad actor is run on the cluster; otherwise, e.g. when it is run during unit tests, set it to True.

Examples

The following actor simulates the performance of two VDisks. The first VDisk logs a message of 65536 bytes every 65536 bytes and deletes data that exceeds 1048576 bytes. The second one writes 1024 bytes as messages of 128 bytes every 2147483647 bytes and deletes data that exceeds 2147483647 bytes.

PDiskLogLoad: {
    Tag: 1
    PDiskId: 1
    PDiskGuid: 12345
    DurationSeconds: 60
    Workers: {
        VDiskId: {GroupID: 1 GroupGeneration: 5 Ring: 1 Domain: 1 VDisk: 1}
        MaxInFlight: 1
        SizeIntervalMin: 65536
        SizeIntervalMax: 65536
        BurstInterval: 65536
        BurstSize: 65536
        StorageDuration: 1048576
    }
    Workers: {
        VDiskId: {GroupID: 2 GroupGeneration: 5 Ring: 1 Domain: 1 VDisk: 1}
        MaxInFlight: 1
        SizeIntervalMin: 128
        SizeIntervalMax: 128
        BurstInterval: 2147483647
        BurstSize: 1024
        StorageDuration: 2147483647
    }
    IsWardenlessTest: false
}

The test passes if none of the cluster nodes got overloaded and the status of the PDisk in question is Normal. You can check this using the cluster Embedded UI.