diff options
Diffstat (limited to 'doc/dev/crimson/crimson.rst')
| -rw-r--r-- | doc/dev/crimson/crimson.rst | 480 |
1 files changed, 480 insertions, 0 deletions
diff --git a/doc/dev/crimson/crimson.rst b/doc/dev/crimson/crimson.rst new file mode 100644 index 000000000..cbc20b773 --- /dev/null +++ b/doc/dev/crimson/crimson.rst @@ -0,0 +1,480 @@ +======= +crimson +======= + +Crimson is the code name of ``crimson-osd``, which is the next +generation ``ceph-osd``. It improves performance when using fast network +and storage devices, employing state-of-the-art technologies including +DPDK and SPDK. BlueStore continues to support HDDs and slower SSDs. +Crimson aims to be backward compatible with the classic ``ceph-osd``. + +.. highlight:: console + +Building Crimson +================ + +Crimson is not enabled by default. Enable it at build time by running:: + + $ WITH_SEASTAR=true ./install-deps.sh + $ mkdir build && cd build + $ cmake -DWITH_SEASTAR=ON .. + +Please note, `ASan`_ is enabled by default if Crimson is built from a source +cloned using ``git``. + +.. _ASan: https://github.com/google/sanitizers/wiki/AddressSanitizer + +Testing crimson with cephadm +=============================== + +The Ceph CI/CD pipeline builds containers with +``crimson-osd`` subsitituted for ``ceph-osd``. + +Once a branch at commit <sha1> has been built and is available in +``shaman``, you can deploy it using the cephadm instructions outlined +in :ref:`cephadm` with the following adaptations. + +First, while performing the initial bootstrap, use the ``--image`` flag to +use a Crimson build: + +.. prompt:: bash # + + cephadm --image quay.ceph.io/ceph-ci/ceph:<sha1>-crimson --allow-mismatched-release bootstrap ... + +You'll likely need to supply the ``--allow-mismatched-release`` flag to +use a non-release branch. + +Additionally, prior to deploying OSDs, you'll need enable Crimson to +direct the default pools to be created as Crimson pools. From the cephadm shell run: + +.. prompt:: bash # + + ceph config set global 'enable_experimental_unrecoverable_data_corrupting_features' crimson + ceph osd set-allow-crimson --yes-i-really-mean-it + ceph config set mon osd_pool_default_crimson true + +The first command enables the ``crimson`` experimental feature. Crimson +is highly experimental, and malfunctions including crashes +and data loss are to be expected. + +The second enables the ``allow_crimson`` OSDMap flag. The monitor will +not allow ``crimson-osd`` to boot without that flag. + +The last causes pools to be created by default with the ``crimson`` flag. +Crimson pools are restricted to operations supported by Crimson. +``Crimson-osd`` won't instantiate PGs from non-Crimson pools. + +Running Crimson +=============== + +As you might expect, Crimson does not yet have as extensive a feature set as does ``ceph-osd``. + +object store backend +-------------------- + +At the moment, ``crimson-osd`` offers both native and alienized object store +backends. The native object store backends perform IO using the SeaStar reactor. +They are: + +.. describe:: cyanstore + + CyanStore is modeled after memstore in the classic OSD. + +.. describe:: seastore + + Seastore is still under active development. + +The alienized object store backends are backed by a thread pool, which +is a proxy of the alienstore adaptor running in Seastar. The proxy issues +requests to object stores running in alien threads, i.e., worker threads not +managed by the Seastar framework. They are: + +.. describe:: memstore + + The memory backed object store + +.. describe:: bluestore + + The object store used by the classic ``ceph-osd`` + +daemonize +--------- + +Unlike ``ceph-osd``, ``crimson-osd`` does not daemonize itself even if the +``daemonize`` option is enabled. In order to read this option, ``crimson-osd`` +needs to ready its config sharded service, but this sharded service lives +in the Seastar reactor. If we fork a child process and exit the parent after +starting the Seastar engine, that will leave us with a single thread which is +a replica of the thread that called `fork()`_. Tackling this problem in Crimson +would unnecessarily complicate the code. + +Since supported GNU/Linux distributions use ``systemd``, which is able to +daemonize the application, there is no need to daemonize ourselves. +Those using sysvinit can use ``start-stop-daemon`` to daemonize ``crimson-osd``. +If this is does not work out, a helper utility may be devised. + +.. _fork(): http://pubs.opengroup.org/onlinepubs/9699919799/functions/fork.html + +logging +------- + +``Crimson-osd`` currently uses the logging utility offered by Seastar. See +``src/common/dout.h`` for the mapping between Ceph logging levels to +the severity levels in Seastar. For instance, messages sent to ``derr`` +will be issued using ``logger::error()``, and the messages with a debug level +greater than ``20`` will be issued using ``logger::trace()``. + ++---------+---------+ +| ceph | seastar | ++---------+---------+ +| < 0 | error | ++---------+---------+ +| 0 | warn | ++---------+---------+ +| [1, 6) | info | ++---------+---------+ +| [6, 20] | debug | ++---------+---------+ +| > 20 | trace | ++---------+---------+ + +Note that ``crimson-osd`` +does not send log messages directly to a specified ``log_file``. It writes +the logging messages to stdout and/or syslog. This behavior can be +changed using ``--log-to-stdout`` and ``--log-to-syslog`` command line +options. By default, ``log-to-stdout`` is enabled, and ``--log-to-syslog`` is disabled. + + +vstart.sh +--------- + +The following options aree handy when using ``vstart.sh``, + +``--crimson`` + Start ``crimson-osd`` instead of ``ceph-osd``. + +``--nodaemon`` + Do not daemonize the service. + +``--redirect-output`` + Redirect the ``stdout`` and ``stderr`` to ``out/$type.$num.stdout``. + +``--osd-args`` + Pass extra command line options to ``crimson-osd`` or ``ceph-osd``. + This is useful for passing Seastar options to ``crimson-osd``. For + example, one can supply ``--osd-args "--memory 2G"`` to set the amount of + memory to use. Please refer to the output of:: + + crimson-osd --help-seastar + + for additional Seastar-specific command line options. + +``--cyanstore`` + Use CyanStore as the object store backend. + +``--bluestore`` + Use the alienized BlueStore as the object store backend. This is the default. + +``--memstore`` + Use the alienized MemStore as the object store backend. + +``--seastore`` + Use SeaStore as the back end object store. + +``--seastore-devs`` + Specify the block device used by SeaStore. + +``--seastore-secondary-devs`` + Optional. SeaStore supports multiple devices. Enable this feature by + passing the block device to this option. + +``--seastore-secondary-devs-type`` + Optional. Specify the type of secondary devices. When the secondary + device is slower than main device passed to ``--seastore-devs``, the cold + data in faster device will be evicted to the slower devices over time. + Valid types include ``HDD``, ``SSD``(default), ``ZNS``, and ``RANDOM_BLOCK_SSD`` + Note secondary devices should not be faster than the main device. + +``--seastore`` + Use SeaStore as the object store backend. + +To start a cluster with a single Crimson node, run:: + + $ MGR=1 MON=1 OSD=1 MDS=0 RGW=0 ../src/vstart.sh -n -x \ + --without-dashboard --cyanstore \ + --crimson --redirect-output \ + --osd-args "--memory 4G" + +Here we assign 4 GiB memory and a single thread running on core-0 to ``crimson-osd``. + +Another SeaStore example:: + + $ MGR=1 MON=1 OSD=1 MDS=0 RGW=0 ../src/vstart.sh -n -x \ + --without-dashboard --seastore \ + --crimson --redirect-output \ + --seastore-devs /dev/sda \ + --seastore-secondary-devs /dev/sdb \ + --seastore-secondary-devs-type HDD + +Stop this ``vstart`` cluster by running:: + + $ ../src/stop.sh --crimson + +Metrics and Tracing +=================== + +Crimson offers three ways to report stats and metrics. + +pg stats reported to mgr +------------------------ + +Crimson collects the per-pg, per-pool, and per-osd stats in a `MPGStats` +message which is sent to the Ceph Managers. Manager modules can query +them using the `MgrModule.get()` method. + +asock command +------------- + +An admin socket command is offered for dumping metrics:: + + $ ceph tell osd.0 dump_metrics + $ ceph tell osd.0 dump_metrics reactor_utilization + +Here `reactor_utilization` is an optional string allowing us to filter +the dumped metrics by prefix. + +Prometheus text protocol +------------------------ + +The listening port and address can be configured using the command line options of +`--prometheus_port` +see `Prometheus`_ for more details. + +.. _Prometheus: https://github.com/scylladb/seastar/blob/master/doc/prometheus.md + +Profiling Crimson +================= + +fio +--- + +``crimson-store-nbd`` exposes configurable ``FuturizedStore`` internals as an +NBD server for use with ``fio``. + +In order to use ``fio`` to test ``crimson-store-nbd``, perform the below steps. + +#. You will need to install ``libnbd``, and compile it into ``fio`` + + .. prompt:: bash $ + + apt-get install libnbd-dev + git clone git://git.kernel.dk/fio.git + cd fio + ./configure --enable-libnbd + make + +#. Build ``crimson-store-nbd`` + + .. prompt:: bash $ + + cd build + ninja crimson-store-nbd + +#. Run the ``crimson-store-nbd`` server with a block device. Specify + the path to the raw device, for example ``/dev/nvme1n1``, in place of the created + file for testing with a block device. + + .. prompt:: bash $ + + export disk_img=/tmp/disk.img + export unix_socket=/tmp/store_nbd_socket.sock + rm -f $disk_img $unix_socket + truncate -s 512M $disk_img + ./bin/crimson-store-nbd \ + --device-path $disk_img \ + --smp 1 \ + --mkfs true \ + --type transaction_manager \ + --uds-path ${unix_socket} & + + Below are descriptions of these command line arguments: + + ``--smp`` + The number of CPU cores to use (Symmetric MultiProcessor) + + ``--mkfs`` + Initialize the device first. + + ``--type`` + The back end to use. If ``transaction_manager`` is specified, SeaStore's + ``TransactionManager`` and ``BlockSegmentManager`` are used to emulate a + block device. Otherwise, this option is used to choose a backend of + ``FuturizedStore``, where the whole "device" is divided into multiple + fixed-size objects whose size is specified by ``--object-size``. So, if + you are only interested in testing the lower-level implementation of + SeaStore like logical address translation layer and garbage collection + without the object store semantics, ``transaction_manager`` would be a + better choice. + +#. Create a ``fio`` job file named ``nbd.fio`` + + .. code:: ini + + [global] + ioengine=nbd + uri=nbd+unix:///?socket=${unix_socket} + rw=randrw + time_based + runtime=120 + group_reporting + iodepth=1 + size=512M + + [job0] + offset=0 + +#. Test the Crimson object store, using the custom ``fio`` built just now + + .. prompt:: bash $ + + ./fio nbd.fio + +CBT +--- +We can use `cbt`_ for performance tests:: + + $ git checkout main + $ make crimson-osd + $ ../src/script/run-cbt.sh --cbt ~/dev/cbt -a /tmp/baseline ../src/test/crimson/cbt/radosbench_4K_read.yaml + $ git checkout yet-another-pr + $ make crimson-osd + $ ../src/script/run-cbt.sh --cbt ~/dev/cbt -a /tmp/yap ../src/test/crimson/cbt/radosbench_4K_read.yaml + $ ~/dev/cbt/compare.py -b /tmp/baseline -a /tmp/yap -v + 19:48:23 - INFO - cbt - prefill/gen8/0: bandwidth: (or (greater) (near 0.05)):: 0.183165/0.186155 => accepted + 19:48:23 - INFO - cbt - prefill/gen8/0: iops_avg: (or (greater) (near 0.05)):: 46.0/47.0 => accepted + 19:48:23 - WARNING - cbt - prefill/gen8/0: iops_stddev: (or (less) (near 0.05)):: 10.4403/6.65833 => rejected + 19:48:23 - INFO - cbt - prefill/gen8/0: latency_avg: (or (less) (near 0.05)):: 0.340868/0.333712 => accepted + 19:48:23 - INFO - cbt - prefill/gen8/1: bandwidth: (or (greater) (near 0.05)):: 0.190447/0.177619 => accepted + 19:48:23 - INFO - cbt - prefill/gen8/1: iops_avg: (or (greater) (near 0.05)):: 48.0/45.0 => accepted + 19:48:23 - INFO - cbt - prefill/gen8/1: iops_stddev: (or (less) (near 0.05)):: 6.1101/9.81495 => accepted + 19:48:23 - INFO - cbt - prefill/gen8/1: latency_avg: (or (less) (near 0.05)):: 0.325163/0.350251 => accepted + 19:48:23 - INFO - cbt - seq/gen8/0: bandwidth: (or (greater) (near 0.05)):: 1.24654/1.22336 => accepted + 19:48:23 - INFO - cbt - seq/gen8/0: iops_avg: (or (greater) (near 0.05)):: 319.0/313.0 => accepted + 19:48:23 - INFO - cbt - seq/gen8/0: iops_stddev: (or (less) (near 0.05)):: 0.0/0.0 => accepted + 19:48:23 - INFO - cbt - seq/gen8/0: latency_avg: (or (less) (near 0.05)):: 0.0497733/0.0509029 => accepted + 19:48:23 - INFO - cbt - seq/gen8/1: bandwidth: (or (greater) (near 0.05)):: 1.22717/1.11372 => accepted + 19:48:23 - INFO - cbt - seq/gen8/1: iops_avg: (or (greater) (near 0.05)):: 314.0/285.0 => accepted + 19:48:23 - INFO - cbt - seq/gen8/1: iops_stddev: (or (less) (near 0.05)):: 0.0/0.0 => accepted + 19:48:23 - INFO - cbt - seq/gen8/1: latency_avg: (or (less) (near 0.05)):: 0.0508262/0.0557337 => accepted + 19:48:23 - WARNING - cbt - 1 tests failed out of 16 + +Here we compile and run the same test against two branches: ``main`` and ``yet-another-pr``. +We then compare the results. Along with every test case, a set of rules is defined to check for +performance regressions when comparing the sets of test results. If a possible regression is found, the rule and +corresponding test results are highlighted. + +.. _cbt: https://github.com/ceph/cbt + +Hacking Crimson +=============== + + +Seastar Documents +----------------- + +See `Seastar Tutorial <https://github.com/scylladb/seastar/blob/master/doc/tutorial.md>`_ . +Or build a browsable version and start an HTTP server:: + + $ cd seastar + $ ./configure.py --mode debug + $ ninja -C build/debug docs + $ python3 -m http.server -d build/debug/doc/html + +You might want to install ``pandoc`` and other dependencies beforehand. + +Debugging Crimson +================= + +Debugging with GDB +------------------ + +The `tips`_ for debugging Scylla also apply to Crimson. + +.. _tips: https://github.com/scylladb/scylla/blob/master/docs/dev/debugging.md#tips-and-tricks + +Human-readable backtraces with addr2line +---------------------------------------- + +When a Seastar application crashes, it leaves us with a backtrace of addresses, like:: + + Segmentation fault. + Backtrace: + 0x00000000108254aa + 0x00000000107f74b9 + 0x00000000105366cc + 0x000000001053682c + 0x00000000105d2c2e + 0x0000000010629b96 + 0x0000000010629c31 + 0x00002a02ebd8272f + 0x00000000105d93ee + 0x00000000103eff59 + 0x000000000d9c1d0a + /lib/x86_64-linux-gnu/libc.so.6+0x000000000002409a + 0x000000000d833ac9 + Segmentation fault + +The ``seastar-addr2line`` utility provided by Seastar can be used to map these +addresses to functions. The script expects input on ``stdin``, +so we need to copy and paste the above addresses, then send EOF by inputting +``control-D`` in the terminal. One might use ``echo`` or ``cat`` instead`:: + + $ ../src/seastar/scripts/seastar-addr2line -e bin/crimson-osd + + 0x00000000108254aa + 0x00000000107f74b9 + 0x00000000105366cc + 0x000000001053682c + 0x00000000105d2c2e + 0x0000000010629b96 + 0x0000000010629c31 + 0x00002a02ebd8272f + 0x00000000105d93ee + 0x00000000103eff59 + 0x000000000d9c1d0a + 0x00000000108254aa + [Backtrace #0] + seastar::backtrace_buffer::append_backtrace() at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:1136 + seastar::print_with_backtrace(seastar::backtrace_buffer&) at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:1157 + seastar::print_with_backtrace(char const*) at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:1164 + seastar::sigsegv_action() at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:5119 + seastar::install_oneshot_signal_handler<11, &seastar::sigsegv_action>()::{lambda(int, siginfo_t*, void*)#1}::operator()(int, siginfo_t*, void*) const at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:5105 + seastar::install_oneshot_signal_handler<11, &seastar::sigsegv_action>()::{lambda(int, siginfo_t*, void*)#1}::_FUN(int, siginfo_t*, void*) at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:5101 + ?? ??:0 + seastar::smp::configure(boost::program_options::variables_map, seastar::reactor_config) at /home/kefu/dev/ceph/build/../src/seastar/src/core/reactor.cc:5418 + seastar::app_template::run_deprecated(int, char**, std::function<void ()>&&) at /home/kefu/dev/ceph/build/../src/seastar/src/core/app-template.cc:173 (discriminator 5) + main at /home/kefu/dev/ceph/build/../src/crimson/osd/main.cc:131 (discriminator 1) + +Note that ``seastar-addr2line`` is able to extract addresses from +its input, so you can also paste the log messages as below:: + + 2020-07-22T11:37:04.500 INFO:teuthology.orchestra.run.smithi061.stderr:Backtrace: + 2020-07-22T11:37:04.500 INFO:teuthology.orchestra.run.smithi061.stderr: 0x0000000000e78dbc + 2020-07-22T11:37:04.501 INFO:teuthology.orchestra.run.smithi061.stderr: 0x0000000000e3e7f0 + 2020-07-22T11:37:04.501 INFO:teuthology.orchestra.run.smithi061.stderr: 0x0000000000e3e8b8 + 2020-07-22T11:37:04.501 INFO:teuthology.orchestra.run.smithi061.stderr: 0x0000000000e3e985 + 2020-07-22T11:37:04.501 INFO:teuthology.orchestra.run.smithi061.stderr: /lib64/libpthread.so.0+0x0000000000012dbf + +Unlike the classic ``ceph-osd``, Crimson does not print a human-readable backtrace when it +handles fatal signals like `SIGSEGV` or `SIGABRT`. It is also more complicated +with a stripped binary. So instead of planting a signal handler for +those signals into Crimson, we can use `script/ceph-debug-docker.sh` to map +addresses in the backtrace:: + + # assuming you are under the source tree of ceph + $ ./src/script/ceph-debug-docker.sh --flavor crimson master:27e237c137c330ebb82627166927b7681b20d0aa centos:8 + .... + [root@3deb50a8ad51 ~]# wget -q https://raw.githubusercontent.com/scylladb/seastar/master/scripts/seastar-addr2line + [root@3deb50a8ad51 ~]# dnf install -q -y file + [root@3deb50a8ad51 ~]# python3 seastar-addr2line -e /usr/bin/crimson-osd + # paste the backtrace here |