Home >Backend Development >PHP Tutorial >Extension development of ceph management platform Calamari_PHP tutorial
Extension development of ceph management platform Calamari_PHP tutorial
- WBOYOriginal
- 2016-07-12 09:00:371210browse
Extended development of the ceph management platform Calamari
I haven’t written a log for nearly half a year, maybe because I am getting lazy. But sometimes writing can help you settle down, so I might as well come back and record it. I have been working for more than half a year and have become familiar with some related work. I am currently mainly engaged in the research and development of distributed systems. The current development is mainly at the management level and has not yet reached the point of modifying the code. In the past six months, I have become familiar with two very good distributed systems, glusterfs and Ceph. Both distributed storage products have their own advantages, among which Glusterfs provides file services that the Ceph system cannot provide. Ceph's unified architecture of block devices, object storage, and file systems cannot be satisfied by GlusterFs. So each has its own advantages.From a code level, the code of GlusterFs is relatively simple, the layers are obvious, and the stack processing flow is very clear. It is very easy to expand the function of the file system (just add processing modules on the client and server). Although the server-side and client-side codes are one code, the overall code is relatively clear and the amount of code is small.
Ceph is developed in C, and the system itself has multiple processes. Multiple processes form a large cluster, and there are also small clusters inside the cluster. Compared with Glusterfs, the code is much more complicated. Ceph itself realizes self-adjustment and self-healing. Supports the customization of software systems and finds the storage location of objects through the Crush algorithm.
As far as the current popularity is concerned, Ceph is relatively popular, but for the file system provided, Glusterfs is still a good choice.
I have recently been engaged in the development of Ceph-related management platforms and have become familiar with the officially provided Calamari platform. This platform currently mainly provides management of Ceph distributed storage systems. Overall, it mainly provides page management of Ceph. means. From the current implementation perspective, the platform still has certain limitations and cannot complete powerful functions, or the currently provided version can only provide some basic functions. But the Calamari framework is really good. Ceph is open source software, Calamari is also open source software, and Calamari is a combination of a series of open source software. These open source software only complete their specific functions. Although it is pieced together, overall, the framework of this management platform is worth learning from.
The following parts refer to http://www.openstack.cn/?p=2708.
Calamari architecture diagramThe red box part is the part implemented by Calamari code, and the non-red box part is the open source framework that is not implemented by Calamari.
The components installed on the Cephserver node include Diamond and Salt-minion. Diamond is responsible for collecting monitoring data. It supports many data types and metrics. Each type of data is a collector in the above figure. In addition to collecting the status information of Ceph itself, it can also collect key resource usage and performance. Data, including CPU, memory, network, I/O load and disk metrics. Collector uses the local command line to collect data and then reports it to Graphite.
Graphite is not only an enterprise-level monitoring tool, it can also draw in real time. carbon-cache is a back-end process of a highly scalable event-driven I/O architecture implemented in Python. It can effectively communicate with a large number of clients and handle a large amount of business volume with low overhead.
Whisper is similar to RRDtool, providing a database development library for applications to manipulate and retrieve file data (time data point data) stored in special formats. The most basic operation of Whisper is to create new Whisper files and update them. Write new data points to a file and get the retrieved data points
Graphite_web is a user interface used to generate images. Users can access these generated images directly through URLs.
Calamari uses Saltstack to communicate between Calamari Server and Ceph server node. Saltstack is an open source automated operation and maintenance management tool with similar functions to Chef and Puppet. Salt-master sends instructions to the designated Salt-minion to complete the management of Cpeh Cluster; Salt-minion will synchronize and install a ceph.py file from the master after the Ceph server node is installed, which contains the API for Ceph operations. It will Call librados or the command line to finally communicate with Ceph Cluster.
calamari_rest provides Calamari REST API. Please refer to the official documentation for detailed interfaces. Ceph's REST API is a low-level interface, in which each URL maps directly to the equivalent CEPH CLI; Calamari REST API provides a higher-level interface, and API users can use GET/POST/PATCH. method to operate objects without knowing the underlying Ceph commands; the main difference between them is that users of Ceph's REST API need to know Ceph itself very well, while Calamari's REST API is closer to the description of Ceph resources, so it is more Suitable for calling upper-layer applications.
cthulhu can be understood as the Service layer of Calamari Server. It provides an interface for API on the upper side and calls Salt-master on the lower side.
calamari_clients is a set of user interfaces. During the installation process, Calamari Server will first create the opt/calamari/webapp directory, and check in the manager.py (django configuration) file under webapp/calamari. calamari_web All content must be placed under opt/calamari/webapp to provide the UI access page.
The files under the calamari-web package provide all web-related configurations, both calamari_rest and calamari_clients are used.This framework uses a lot of open source software, but it is worth learning from an expansion perspective. Saltstack implements the communication link between the management node and the server node, and supports the management of multiple nodes, so it is not The communication problem between the management node and the server needs to be considered. On the server side, only specific business logic needs to be implemented, that is, the implementation of specific management tasks. At the same time, Saltstack is developed using Python, which facilitates rapid system development and is very convenient for managers to debug and locate problems on site. Ceph itself also provides a Python API interface, and cluster control can be achieved directly through Ceph's API. The use of SaltStack allows the cluster to reach a certain scale. The Master side of SaltStack actually serves as the control interface of the management side, and SaltStack serves as the Agent side of the server. In Calamari, heartbeat messages are sent through Saltstack, the server information and cluster information are checked, and the distribution of commands is controlled. It can be said that if you understand the basic model of SaltStack, you can understand the development and expansion of Calamari.
Another very important set of open source software in this framework is diamond graphite, in which diamond completes the collection of server-side information, and graphite implements the provision of chart information. Diamond currently provides information collection for most open source systems and provides collection of basic server information (CPU, memory, disk, etc.). It is also implemented in Python and is very easy to expand and debug. Currently, Ceph information collection already exists in diamond. Graphite mainly provides time series data for the front desk, which simplifies the rewriting of specific business logic.
To learn and understand Calamari, you must understand some basic components and master the functions and purposes of these components. The following describes how to extend Calamari from the code level.
1 Calamari extension
Develop new functions based on Calamari, which are mainly divided into the following modules. This part includes the Rest-API part, Cthulhu, and salt client Extension. The basic steps for extending new functions are as follows:
>> Expand the URL module and determine the corresponding response interface parameters and the corresponding response interface in the ViewSet.
>> Complete the implementation of some interfaces in ViewSet. This part mainly involves interaction with cthulhu, how to obtain data information, and in some cases, it is also necessary to obtain the serialization operation of the object in the serializer.
>> Complete the expansion of the corresponding type in background rpc.py. This part is mainly for some post operations.
>> Complete the expansion of cluster_monitor.py. Some functions that provide operations need to support create, update, delete and other operations, and the corresponding RequestFactory must be provided. In cluster_monitor.py, the corresponding RequestFactory needs to be added to the code.
>> Complete the writing of the corresponding RequestFactory class. This part is mainly to complete the encapsulation of command operations. And build the corresponding request operation.
>> Extension of salt-minion, this part is mainly for the extension of ceph.py file, of course, new xxx.py file can also be provided.
The following will take the control and operation of PG as an example.
1.1URL module extension
Currently Calmamari adopts Rest-API form and is supported by Django’s Rest-Framework framework. This part is in the rest-api code directory. Django adopts the implementation method of separating Url and code logic, so the URL can be expanded independently.
Add the following PG-related URL to rest-api/calamari-rest/urls/v2.py:
url(r'^cluster/(?P
url(r'^cluster/(?P
calamari_rest.views.v2.PgViewSet.as_view({'post': 'apply'}),
name='cluster-pool-pg-control'),
Two URLs are defined above, which are:
api/v2/cluster/xxxx/pool/x/pg
api/v2/cluster/xxxx/pool/x/pg/ xx/command/xxx
The above two URLs respectively specify the interfaces in PgViewSet, and the get method of the URL corresponds to the list interface. The apply interface corresponding to the post interface. These two interfaces must be implemented in PgViewSet.
1.2 Extension of ViewSet
After extending the URL, the next step is to extend the corresponding response interface. This part of the extension is mainly to implement the interface class specified in the URL. In the previous PG, two different interfaces were specified, namely acquisition and operation commands. The corresponding code path is /rest-api/calamari-rest/view/v2.py. The specific code is as follows:
class PgViewSet(RPCViewSet):
serializer_class= PgSerializer
deflist(self, request, fsid, pool_id):
poolName = self.client. get(fsid, POOL, int(pool_id))['pool_name']
pg_summary = self.client.get_sync_object(fsid, PgSummary.str)
pg_pools = pg_summary['pg_pools'] ['by_pool'][int(pool_id)]
forpg in pg_pools:
pg['pool'] = poolName
return Response( PgSerializer(pg_pools, many=True).data)
defapply(self, request, fsid, pool_id, pg_id, command):
return Response(self.client .apply(fsid, PG, pg_id, command), status=202)
As can be seen from the above implementation, the code implements two interfaces, namely the list and apply interfaces, which correspond to the previous get and post operate. The above two operations will interact with the background cthulhu. They are to obtain parameters and submit requests. There are also certain differences in the returned content.
At the same time, serialization settings are made in the list interface, namely PgSerializer, which is implemented in rest-api/calamari-rest/serializer/v2.py.
1.2.1 Serialization operation
Usually data is serialized in Rest-Api. This part is not necessary. It is usually necessary in operations that need to be changed. of. The following is the serialization operation of Pg:
class PgSerializer(serializers.Serializer):
classMeta:
fields = ('id', 'pool', 'state' , 'up', 'acting', 'up_primary','acting_primary')
id =serializers.CharField(source='pgid')
pool =serializers.CharField(help_text=' pool name')
state =serializers.CharField(source='state', help_text='pg state')
up =serializers.Field(help_text='pg Up set')
acting =serializers.Field(help_text='pg acting set')
up_primary = serializers.IntegerField(help_text='pg up primary')
acting_primary =serializers.IntegerField (help_text='pg acting primary')
This part is not necessary. Some modules may not have this part of the operation. In the previous three steps, the extension of the Rest-API part was basically implemented, among which the main extension was the ViewSet. The relevant ViewSet actually implements the interaction method between cthulhu and rest-api.
In the extension of ViewSet, rpc is actually used to interact with the background, so the implementation part of cthulhu mainly handles the corresponding rpc requests.
1.3rpc extension
rpc.py implements all requested operations, but new extended operations also need to support extensions. Take pg as an example to continue the explanation:
defapply(self, fs_id, object_type, object_id, command):
"""
Apply commands that do not modify an object in a cluster.
"""
cluster = self._fs_resolve(fs_id)
ifobject_type == OSD:
# Run a resolve to throw exception if it's unknown
self._osd_resolve(cluster, object_id)
return cluster.request_apply(OSD, object_id, command)
elifobject_type == PG:
return cluster.request_apply(PG ,object_id, command)
else:
raise NotImplementedError(object_type)
The list of Pg is obtained through PgSummary. This part already existed in the previous implementation. The previous code was implemented as follows:
defget_sync_object(self, fs_id, object_type, path=None):
"""
Getone of the objects that ClusterMonitor keeps a copy of from the mon, such
as the cluster maps.
:param fs_id: The fsid of a cluster
:param object_type: String, one of SYNC_OBJECT_TYPES
:param path: List, optional, a path within the object to return instead of the whole thing
:return : the requested data, or None if it was not found (including ifany element of ``path``
was not found)
"""
ifpath:
obj =self._fs_resolve(fs_id).get_sync_object(SYNC_OBJECT_STR_TYPE[object_type])
try:
for part in path:
if isinstance(obj, dict):
obj = obj[part]
else:
obj = getattr(obj, part)
except (AttributeError, KeyError) as e:
log.exception("Exception %s traversing %s: obj=%s" % (e, path,obj))
raise NotFound( object_type, path)
return obj
else:
returnself._fs_resolve(fs_id).get_sync_object_data(SYNC_OBJECT_STR_TYPE[object_type])
1.4cluster_monitor.py Extension
All requested operations will be controlled by the cluster. This part can be implemented through cluster_monitor, taking pg as an example.
def__init__(self, fsid, cluster_name, notifier, persister, servers, eventer, requests):
super(ClusterMonitor, self).__init__()
self.fsid = fsid
self.name = cluster_name
self.update_time = datetime.datetime.utcnow().replace(tzinfo=utc)
self._notifier = notifier
self._persister= persister
self._servers = servers
self._eventer = eventer
self ._requests = requests
#Which mon we are currently using for running requests,
#identified by minion ID
self._favorite_mon = None
self._last_heartbeat = {}
self._complete = gevent.event.Event()
self.done = gevent.event.Event( )
self._sync_objects = SyncObjects(self.name)
self._request_factories = {
CRUSH_MAP: CrushRequestFactory,
OSD: OSDREQUESTFACTORY,
PoolRequestFactory,
CacheTierRequ ESTFactory,
PG: PGREQUESTFACTORY,
Async_command: AsyncComrequestFactory
itor = pluginmonitor (servers)🎜>self._ready = gevent.event.Event()This part is mainly to bind the corresponding request to the corresponding request factory class, so that a suitable request can be generated.
1.5 Factory class writing
This factory class is mainly designed to implement specific interface classes for different needs. Different objects have different request classes. Take Pg as an example:
from cthulhu.manager.request_factory importRequestFactory
from cthulhu.manager.user_request importRadosRequest
from calamari_common.types importPG_IMPLEMENTED_COMMANDS, PgSummary
class PgRequestFactory (RequestFactory) :
def scrub(self,pg_id):
return RadosRequest(
"Initiating scrub on{cluster_name}-pg{id}".format(cluster_name=self. _cluster_monitor.name,id=pg_id),
self._cluster_monitor.fsid,
self._cluster_monitor.name,
[('pg scrub', {'pgid' : pg_id})])
defdeep_scrub(self, pg_id):
return RadosRequest(
"Initiating deep-scrub on{cluster_name}- osd.{id}".format(cluster_name=self._cluster_monitor.name,id=pg_id),
self._cluster_monitor.fsid,
self._cluster_monitor.name,
[('pg deep-scrub', {'pgid': pg_id})])
defrepair(self, pg_id):
return RadosRequest(
"Initiating repair on{cluster_name}-osd.{id}".format(cluster_name=self._cluster_monitor.name,id=pg_id),
self._cluster_monitor.fsid,
self._cluster_monitor.name,
[('pg repair', {'pgid': pg_id})])
defget_valid_commands(self, pg_id) :
ret_val = {}
file('/tmp/pgsummary.txt', 'a ').write(PgSummary.str 'n')
pg_summary = self._cluster_monitor.get_sync_object(PgSummary)
pg_pools = pg_summary['pg_pools']['by_pool']
pool_id = int(pg_id.split('.')[0])
pool= pg_pools[pool_id]
forpg in pool:
if pg['pgid'] == pg_id:
ret_val[pg_id] = {'valid_commands': PG_IMPLEMENTED_COMMANDS}
else:
ret_val[pg_id] = {'valid_commands': []}
return ret_val
This class implements the implementation of three different commands. This command is mainly for corresponding encapsulation. These keywords need to be selected according to the parameters in the ceph source code, so they need to be encoded. Refer to the json parameter name of the corresponding command in the ceph source code.
Extension of 1.6 salt-minion
This part is the extension module of salt, which is mainly used to obtain the corresponding data information, execute the corresponding operation commands, etc. Execute the corresponding operation command through salt in cthulhu. Ceph.py has interfaces such as rados.commands, which can be used to execute ceph commands. The commands encapsulated in the factory class will eventually be executed through this interface.Summary
Overall, Calamari’s code structure is relatively clear, and the open source framework is also worth learning. In subsequent distributed management systems, you can also refer to the architecture of saltstack diamond graphite. The former implements control Logic, the latter two implement data collection and data storage and display.