diff options
Diffstat (limited to 'docs/testing/user')
| -rwxr-xr-x | docs/testing/user/installation.rst | 38 | ||||
| -rw-r--r-- | docs/testing/user/introduction.rst | 39 | ||||
| -rw-r--r-- | docs/testing/user/test-usage.rst | 229 |
3 files changed, 279 insertions, 27 deletions
diff --git a/docs/testing/user/installation.rst b/docs/testing/user/installation.rst index c129fee..7f56244 100755 --- a/docs/testing/user/installation.rst +++ b/docs/testing/user/installation.rst @@ -8,11 +8,20 @@ StorPerf Installation Guide OpenStack Prerequisites =========================== -If you do not have an Ubuntu 16.04 image in Glance, you will need to add one. -You also need to create the StorPerf flavor, or choose one that closely -matches. For Ubuntu 16.04, it must have a minimum of a 4 GB disk. It should -also have about 8 GB RAM to support FIO's memory mapping of written data blocks -to ensure 100% coverage of the volume under test. +StorPerf can be instructed to use OpenStack APIs in order to manage a +Heat stack of virtual machines and Cinder volumes, or it can be run in +stackless mode, where it does not need to know anything about OpenStack. + +When running in OpenStack mode, there will need to be an external network +with floating IPs available to assign to the VMs, as well as a Glance image +that can be used to boot the VMs. This can be almost any Linux based +image, as long as it can either accept OpenStack metadata for injecting +the SSH key, or it has known SSH credentials as part of the base image. + +The flavor for the image should provide enough disk space for the initial +boot, along with additional space if profiling of the Glance backing is +desired. It should also provide at least 8 GB RAM to support FIO's memory +mapping of written data blocks. There are scripts in storperf/ci directory to assist, or you can use the follow code snippets: @@ -34,9 +43,10 @@ code snippets: OpenStack Credentials ~~~~~~~~~~~~~~~~~~~~~ -You must have your OpenStack Controller environment variables defined and passed to -the StorPerf container. The easiest way to do this is to put the rc file contents -into a clean file called admin.rc that looks similar to this for V2 authentication: +Unless running in stackless mode, the OpenStack Controller environment +variables must be defined and passed to the StorPerf container. The easiest +way to do this is to put the rc file contents into a clean file called +admin.rc that looks similar to this for V2 authentication: .. code-block:: console @@ -89,7 +99,7 @@ Requirements: * Host has access to the OpenStack Controller API * Host must have internet connectivity for downloading docker image * Enough OpenStack floating IPs must be available to match your agent count -* A local directory for holding the Carbon DB Whisper files +* Optionally, a local directory for holding the Carbon DB Whisper files Local disk used for the Carbon DB storage as the default size of the docker container is only 10g. Here is an example of how to create a local storage @@ -117,7 +127,7 @@ http://storperf:5000/graphite Running StorPerf Container ========================== -**As of Euphrates (development) release (June 2017), StorPerf has +**As of Euphrates release (June 2017), StorPerf has changed to use docker-compose in order to start its services.** Docker compose requires a local file to be created in order to define the @@ -146,8 +156,12 @@ which should result in: To run, you must specify two environment variables: -* ENV_FILE, which points to your OpenStack admin.rc as noted above. -* CARBON_DIR, which points to a directory that will be mounted to store the raw metrics. +* ENV_FILE, which points to your OpenStack admin.rc as noted above. If running + in stackless mode only, it is possible to remove the ENV_FILE reference from + the docker-compose.yaml file. +* CARBON_DIR, which points to a directory that will be mounted to store the + raw metrics. If desired, the CARBON_DIR can be removed from the + docker-compose.yaml file, causing metrics to be kept in the container only. * TAG, which specified the Docker tag for the build (ie: latest, danube.3.0, etc). The following command will start all the StorPerf services: diff --git a/docs/testing/user/introduction.rst b/docs/testing/user/introduction.rst index 0099c39..c864edc 100644 --- a/docs/testing/user/introduction.rst +++ b/docs/testing/user/introduction.rst @@ -27,16 +27,16 @@ How Does StorPerf Work? Once launched, StorPerf presents a ReST interface, along with a `Swagger UI <https://swagger.io/swagger-ui/>`_ that makes it easier to -form HTTP ReST requests. Issuing an HTTP POST to the configurations API -causes StorPerf to talk to OpenStack's heat service to create a new stack -with as many agent VMs and attached Cinder volumes as specified. +form HTTP ReST requests. -After the stack is created, we can issue one or more jobs by issuing a POST -to the jobs ReST API. The job is the smallest unit of work that StorPerf -can use to measure the disk's performance. +StorPerf enables us to run FIO on multiple VMs, containers or bare +metal servers by providing a recent release of FIO, copying it to the +target system and running I/O workloads specified. It also provides a +simple API to initialize the target device and fill it with random data +to ensure that performance is measured against real data, not blank media. -While the job is running, StorPerf collects the performance metrics from each -of the disks under test every minute. Once the trend of metrics match the +While an FIO job is running, StorPerf collects the performance metrics from +each of the jobs every minute. Once the trend of metrics match the criteria specified in the SNIA methodology, the job automatically terminates and the valid set of metrics are available for querying. @@ -45,6 +45,29 @@ measured start to "flat line" and stay within that range for the specified amount of time, then the metrics are considered to be indicative of a repeatable level of performance. +With OpenStack Heat +~~~~~~~~~~~~~~~~~~~ + +StorPerf provides an API to interact with OpenStack Heat to automatically +create a set of target VMs and Cinder volumes. The Configurations API is +used to specify how many VMs and volumes to create, as well as the size of +each Cinder volume. + +Without OpenStack Heat +~~~~~~~~~~~~~~~~~~~~~~ + +StorPerf can also use IP addresses or DNS names to connect to systems that +have already been provisioned by any external provider, including OpenStack. +By specifying a stack name of 'null' in the JSON payload, StorPerf will look +for a list of IP addresses and credentials to use in order to SSH to the +target systems. In this way, StorPerf can be used to profile bare metal, +containers that have SSH enabled, or VMs running under OpenStack, WMware ESXi, +VIO, Microsoft Hyper-V, or anything else. The only requirement is that +the target be capable of accepting and authenticating SSH connections, and that +it is Linux based, as currently the FIO supplied by StorPerf is not compiled +to run under Microsoft Windows or other non-Linux operating systems. + + StorPerf Testing Guidelines =========================== diff --git a/docs/testing/user/test-usage.rst b/docs/testing/user/test-usage.rst index 41cbbbd..5dfd048 100644 --- a/docs/testing/user/test-usage.rst +++ b/docs/testing/user/test-usage.rst @@ -31,8 +31,15 @@ The typical test execution follows this pattern: #. Execute one or more performance runs #. Delete the environment -Configure The Environment -========================= +OpenStack or Stackless +====================== +StorPerf provides the option of controlling the OpenStack environment +via a Heat Stack, or it can run in stackless mode, where it connects +directly to the IP addresses supplied, regardless of how the slave +was created or even if it is an OpenStack VM. + +Configure The Environment for OpenStack Usage +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The following pieces of information are required to prepare the environment: @@ -51,12 +58,12 @@ The following pieces of information are required to prepare the environment: VMs from properly attaching Cinder volumes. There are two known workarounds: #. Create the environment with 0 Cinder volumes attached, and after the VMs - have finished booting, modify the stack to have 1 or more Cinder volumes. - See section on Changing Stack Parameters later in this guide. + have finished booting, modify the stack to have 1 or more Cinder volumes. + See section on Changing Stack Parameters later in this guide. #. Add the following image metadata to Glance. This will cause the Cinder - volume to be mounted as a SCSI device, and therefore your target will be - /dev/sdb, etc, instead of /dev/vdb. You will need to specify this in your - warm up and workload jobs. + volume to be mounted as a SCSI device, and therefore your target will be + /dev/sdb, etc, instead of /dev/vdb. You will need to specify this in your + warm up and workload jobs. .. code-block: --property hw_disk_bus=scsi --property hw_scsi_model=virtio-scsi @@ -83,6 +90,26 @@ takes a JSON payload as follows. This call will block until the stack is created, at which point it will return the OpenStack heat stack id as well as the IP addresses of the slave agents. + +Configure The Environment for Stackless Usage +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +To configure the environment for stackless usage, the slaves must be +fully operational (ie: a Linux operating system is running, are reachable +via TCP/IP address or hostname). + +It is not necessary to use the Configurations API, but instead define the +stack name as 'null' in any of the other APIs. This instructs StorPerf not to +gather information about the stack from OpenStack, and to simply use the +supplied IP addresses and credentials to communicate with the slaves. + +A slave can be a container (provided we can SSH to it), a VM running in any +hypervisor, or even a bare metal server. In the bare metal case, it even +allows for performing RADOS or RDB performance tests using the appropriate +FIO engine. + + + Initialize the Target Volumes ============================= Before executing a test run for the purpose of measuring performance, it is @@ -120,6 +147,137 @@ This will return a job ID as follows. This job ID can be used to query the state to determine when it has completed. See the section on querying jobs for more information. +Authentication and Slave Selection +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +It is possible to run the Initialization API against a subset of the slaves +known to the stack, or to run it in stackless mode, where StorPerf +connects directly to the IP addresses supplied via SSH. The following +keys are available: + +slave_addresses + (optional) A list of IP addresses or hostnames to use as targets. If + omitted, and StorPerf is not running in stackless mode, the full list of + IP addresses from the OpenStack Heat stack is used. + +stack_name + (optional) Either the name of the stack in Heat to use, or null if running + in stackless mode. + +username + (optional) The username to supply to SSH when logging in. This defaults to + 'storperf' if not supplied. + +password + (optional) The password to supply to SSH when logging in. If omitted, the + SSH key is used instead. + +ssh_private_key + (optional) The SSH private key to supply to SSH when logging in. If omitted, + the default StorPerf private key is used. + +This shows an example of stackless mode going against a single bare metal +server reachable by IP address: + +.. code-block:: json + + { + "username": "labadmin", + "ssh_private_key": "-----BEGIN RSA PRIVATE KEY----- \nMIIE...X0=\n-----END RSA PRIVATE KEY-----", + "slave_addresses": [ + "172.17.108.44" + ], + "stack_name": null, + } + + +Filesystems and Mounts +~~~~~~~~~~~~~~~~~~~~~~ + +It is also possible to instruct StorPerf to create a file system on a device +and mount that as the target directory. The filesystem can be anything +supported by the target slave OS and it is possible to pass specific arguments +to the mkfs command. The following additional keys are available in the +Initializations API for file system control: + +mkfs + The type and arguments to pass for creating a filesystem + +mount_device + The target device on which to make the file system. The file system will + be mounted on the target specified. + +The following example shows the forced creation (-f) of an XFS filesystem +on device /dev/sdb, and mounting that device on /storperf/filesystem. + +**Note** If any of the commands (mkfs, mount) fail for any reason, the +Initializations API will return with a 400 code and the body of the response +will contain the error message. + +.. code-block:: json + + { + "target": "/storperf/filesystem", + "mkfs": "xfs -f", + "mount_device": "/dev/sdb", + } + + +Initializing Filesystems +~~~~~~~~~~~~~~~~~~~~~~~~ + +Just like we need to fill Cinder volumes with data, if we want to profile +files on a mounted file system, we need to initialize the file sets with +random data prior to starting a performance run. The Initializations API +can also be used to create test data sets. + +**Note** be sure to use the same parameters for the number of files, sizes +and jobs in both the Initializations API and the Jobs API, or you will end +up with possibly incorrect results in the Job performance run. + +The following keys are available in the Initializations API for file creation: + +filesize + The size of each file to be created and filled with random data. + +nrfiles + The number of files per job to create. + +numjobs + The number of independent instances of FIO to launch. + +Example: + +.. code-block:: json + + { + "target": "/storperf/filesystem", + "filesize": "2G", + "nrfiles": 10, + "numjobs": 10 + } + +This would create 100 (10 nrfiles x 10 numjobs) 2G files in the directory +/storperf/filesystem. + + +.. code-block:: json + + { + "username": "labadmin", + "ssh_private_key": "-----BEGIN RSA PRIVATE KEY----- \nMIIE...X0=\n-----END RSA PRIVATE KEY-----", + "slave_addresses": [ + "172.17.108.44" + ], + "stack_name": null, + "target": "/storperf/filesystem", + "mkfs": "ext4", + "mount_device": "/dev/sdb", + "filesize": "2G", + "nrfiles": 10, + "numjobs": 10 + } + + Execute a Performance Run ========================= Performance runs can execute either a single workload, or iterate over a matrix @@ -221,6 +379,63 @@ StorPerf will also do a verification of the arguments given prior to returning a Job ID from the ReST call. If an argument fails validation, the error will be returned in the payload of the response. +File System Profiling +~~~~~~~~~~~~~~~~~~~~~ + +As noted in the Initializations API, files in a file system should be +initialized prior to executing a performance run, and the number of jobs, +files and size of files should match the initialization. Given the following +Initializations API call: + +.. code-block:: json + + { + "username": "labadmin", + "ssh_private_key": "-----BEGIN RSA PRIVATE KEY----- \nMIIE...X0=\n-----END RSA PRIVATE KEY-----", + "slave_addresses": [ + "172.17.108.44" + ], + "stack_name": null, + "target": "/storperf/filesystem", + "mkfs": "ext4", + "mount_device": "/dev/sdb", + "filesize": "2G", + "nrfiles": 10, + "numjobs": 10 + } + +The corresponding call to the Jobs API would appear as follows: + +.. code-block:: json + + { + "username": "labadmin", + "ssh_private_key": "-----BEGIN RSA PRIVATE KEY----- \nMIIE...X0=\n-----END RSA PRIVATE KEY-----", + "slave_addresses": [ + "172.17.108.44" + ], + "stack_name": null, + "target": "/storperf/filesystem", + "block_sizes": "4k", + "queue_depths": "8", + "workloads": { + "readwritemix": { + "rw": "rw", + "filesize": "2G", + "nrfiles": "10", + "numjobs": "10" + } + } + } + +**Note** the queue depths and block sizes as well as the I/O pattern (rw) +can change, but the filesize, nrfiles, numjobs and slave addresses must +match the initialization or the performance run could contain skewed results +due to disk initialization. StorPerf explicitly allows for the mismatch +of these so that it is possible to visualize performance when the files +or disks have not been properly initialized. + + Block Sizes ~~~~~~~~~~~ A comma delimited list of the different block sizes to use when reading and |