ceph-rgw zipper的设计理念（2）

本文简介

书接上文。本文以CreateBucket为例进行详细讲述设计理念以及接口变化趋势。

1、接收请求和协议处理请求

 rgw_asio_frontend.cc

主要功能：回调函数注册和请求处理

void handle_connection(boost::asio::io_context& context,
                       RGWProcessEnv& env, Stream& stream,
                       timeout_timer& timeout, size_t header_limit,
                       parse_buffer& buffer, bool is_ssl,
                       SharedMutex& pause_mutex,
                       rgw::dmclock::Scheduler *scheduler,
                       const std::string& uri_prefix,
                       boost::system::error_code& ec,
                       spawn::yield_context yield)
{
  // don't impose a limit on the body, since we read it in pieces
  static constexpr size_t body_limit = std::numeric_limits<size_t>::max();

  auto cct = env.driver->ctx();

  // read messages from the stream until eof
  for (;;) {
    // configure the parser
    rgw::asio::parser_type parser;
    parser.header_limit(header_limit);
    parser.body_limit(body_limit);
    timeout.start();
    // parse the header
    http::async_read_header(stream, buffer, parser, yield[ec]);
    timeout.cancel();
    if (ec == boost::asio::error::connection_reset ||
        ec == boost::asio::error::bad_descriptor ||
        ec == boost::asio::error::operation_aborted ||
#ifdef WITH_RADOSGW_BEAST_OPENSSL
        ec == ssl::error::stream_truncated ||
#endif
        ec == http::error::end_of_stream) {
      ldout(cct, 20) << "failed to read header: " << ec.message() << dendl;
      return;
    }
    auto& message = parser.get();

    bool expect_continue = (message[http::field::expect] == "100-continue");

    {
      // process the request
      RGWRequest req{env.driver->get_new_req_id()};

      StreamIO real_client{cct, stream, timeout, parser, yield, buffer,
                           is_ssl, local_endpoint, remote_endpoint};

      auto real_client_io = rgw::io::add_reordering(
                              rgw::io::add_buffering(cct,
                                rgw::io::add_chunking(
                                  rgw::io::add_conlen_controlling(
                                    &real_client))));
      RGWRestfulIO client(cct, &real_client_io);
      optional_yield y = null_yield;

      process_request(env, &req, uri_prefix, &client, y,
                      scheduler, &user, &latency, &http_ret);
    }

    if (!parser.keep_alive()) {
      return;
    }

    // if we failed before reading the entire message, discard any remaining
    // bytes before reading the next
    while (!expect_continue && !parser.is_done()) {
      static std::array<char, 1024> discard_buffer;

      auto& body = parser.get().body();
      body.size = discard_buffer.size();
      body.data = discard_buffer.data();

      timeout.start();
      http::async_read_some(stream, buffer, parser, yield[ec]);
      timeout.cancel();
      if (ec == http::error::need_buffer) {
        continue;
      }
      if (ec == boost::asio::error::connection_reset) {
        return;
      }
      if (ec) {
        ldout(cct, 5) << "failed to discard unread message: "
            << ec.message() << dendl;
        return;
      }
    }
  }
}

rgw_process.cc

主要功能：请求处理，包括身份认证、请求处理，函数调用返回等。

int process_request(const RGWProcessEnv& penv,
                    RGWRequest* const req,
                    const std::string& frontend_prefix,
                    RGWRestfulIO* const client_io,
                    optional_yield yield,
            rgw::dmclock::Scheduler *scheduler,
                    string* user,
                    ceph::coarse_real_clock::duration* latency,
                    int* http_ret)
{
  int ret = client_io->init(g_ceph_context);
  dout(1) << "====== starting new request req=" << hex << req << dec
      << " =====" << dendl;
  perfcounter->inc(l_rgw_req);

  RGWEnv& rgw_env = client_io->get_env();

  req_state rstate(g_ceph_context, penv, &rgw_env, req->id);
  req_state *s = &rstate;
  rgw::sal::Driver* driver = penv.driver;


  RGWHandler_REST *handler = rest->get_handler(driver, s,
                                               *penv.auth_registry,
                                               frontend_prefix,
                                               client_io, &mgr, &init_error);

  ldpp_dout(s, 2) << "getting op " << s->op << dendl;
  op = handler->get_op();

  std::tie(ret,c) = schedule_request(scheduler, s, op);

  req->op = op;
  ldpp_dout(op, 10) << "op=" << typeid(*op).name() << dendl;
  s->op_type = op->get_type();

  try {
    ldpp_dout(op, 2) << "verifying requester" << dendl;
    ret = op->verify_requester(*penv.auth_registry, yield);
    ldpp_dout(op, 2) << "normalizing buckets and tenants" << dendl;
    ret = handler->postauth_init(yield);

    ret = rgw_process_authenticated(handler, op, req, s, yield, driver);

  } catch (const ceph::crypto::DigestException& e) {
    dout(0) << "authentication failed" << e.what() << dendl;
    abort_early(s, op, -ERR_INVALID_SECRET_KEY, handler, yield);
  }

done:
  try {
    client_io->complete_request();
  } catch (rgw::io::Exception& e) {
    dout(0) << "ERROR: client_io->complete_request() returned "
            << e.what() << dendl;
  }

  if (handler)
    handler->put_op(op);
  rest->put_handler(handler);

  const auto lat = s->time_elapsed();
  if (latency) {
    *latency = lat;
  }
  dout(1) << "====== req done req=" << hex << req << dec
      << " op status=" << op_ret
      << " http_status=" << s->err.http_ret
      << " latency=" << lat
      << " ======"
      << dendl;

  return (ret < 0 ? ret : s->err.ret);
} /* process_request */

在rgw_process_authenticated函数中进行OP的详细处理。包括身份认证、pre-exec、exec、complete等函数。

int rgw_process_authenticated(RGWHandler_REST * const handler,
                              RGWOp *& op,
                              RGWRequest * const req,
                              req_state * const s,
                                    optional_yield y,
                              rgw::sal::Driver* driver,
                              const bool skip_retarget)
{
  ldpp_dout(op, 2) << "init permissions" << dendl;
  int ret = handler->init_permissions(op, y);
  ldpp_dout(op, 2) << "init op" << dendl;
  ret = op->init_processing(y);

  ldpp_dout(op, 2) << "verifying op mask" << dendl;
  ret = op->verify_op_mask();

  ldpp_dout(op, 2) << "verifying op permissions" << dendl;
  {
    auto span = tracing::rgw::tracer.add_span("verify_permission", s->trace);
    std::swap(span, s->trace);
    ret = op->verify_permission(y);
    std::swap(span, s->trace);
  }

  ldpp_dout(op, 2) << "verifying op params" << dendl;
  ret = op->verify_params();
  ldpp_dout(op, 2) << "executing" << dendl;
  {
    auto span = tracing::rgw::tracer.add_span("execute", s->trace);
    std::swap(span, s->trace);
    op->execute(y);
    std::swap(span, s->trace);
  }

  ldpp_dout(op, 2) << "completing" << dendl;
  op->complete();

  return 0;
}

rgw_op.cc

此处忽略rest或者swift中的协议处理过程，直接到RGWOP::createBucket()中

void RGWCreateBucket::execute(optional_yield y)
{
  const rgw::SiteConfig& site = *s->penv.site;
  const std::optional<RGWPeriod>& period = site.get_period();
  const RGWZoneGroup& my_zonegroup = site.get_zonegroup();

  /*步骤1：处理zonegroup信息，确定桶的placement、storage_class等信息，以及是否是主站点存储*/
  /*步骤2：读取桶的信息，如果存在则进行一些处理*/
  // read the bucket info if it exists
  op_ret = driver->load_bucket(this, rgw_bucket(s->bucket_tenant, s->bucket_name),
                               &s->bucket, y);

  /*步骤3：如果桶不存在，则初始化各种信息，*/
  s->bucket_owner.id = s->user->get_id();
  s->bucket_owner.display_name = s->user->get_display_name();
  createparams.owner = s->user->get_id();

  buffer::list aclbl;
  policy.encode(aclbl);
  createparams.attrs[RGW_ATTR_ACL] = std::move(aclbl);

  if (has_cors) {
    buffer::list corsbl;
    cors_config.encode(corsbl);
    createparams.attrs[RGW_ATTR_CORS] = std::move(corsbl);
  }

  /*步骤4：创建桶*/
  ldpp_dout(this, 10) << "user=" << s->user << " bucket=" << s->bucket << dendl;
  op_ret = s->bucket->create(this, createparams, y);
  /*步骤5：如果失败，则回退处理*/
  .....
}

2、store层处理和rados中的处理

int RadosBucket::create(const DoutPrefixProvider* dpp,
                        const CreateParams& params,
                        optional_yield y)
{
  rgw_bucket key = get_key();
  key.marker = params.marker;
  key.bucket_id = params.bucket_id;

  /*创建桶，此处调用rados.cc中的处理流程*/
  int ret = store->getRados()->create_bucket(
      dpp, y, key, params.owner, params.zonegroup_id,
      params.placement_rule, params.zone_placement, params.attrs,
      params.obj_lock_enabled, params.swift_ver_location,
      params.quota, params.creation_time, &bucket_version, info);

  /*link处理*/
  ret = link(dpp, params.owner, y, false);
  if (ret && !existed && ret != -EEXIST) {
    /* if it exists (or previously existed), don't remove it! */
    ret = unlink(dpp, params.owner, y);
    if (ret < 0) {
      ldpp_dout(dpp, 0) << "WARNING: failed to unlink bucket: ret=" << ret
               << dendl;
    }
  } else if (ret == -EEXIST || (ret == 0 && existed)) {
    ret = -ERR_BUCKET_EXISTS;
  }

  return ret;
}
int RGWRados::create_bucket(const DoutPrefixProvider* dpp,
                            optional_yield y,
                            const rgw_bucket& bucket,
                            const rgw_user& owner,
                            const std::string& zonegroup_id,
                            const rgw_placement_rule& placement_rule,
                            const RGWZonePlacementInfo* zone_placement,
                            const std::map<std::string, bufferlist>& attrs,
                            bool obj_lock_enabled,
                            const std::optional<std::string>& swift_ver_location,
                            const std::optional<RGWQuotaInfo>& quota,
                            std::optional<ceph::real_time> creation_time,
                            obj_version* pep_objv,
                            RGWBucketInfo& info)
{
  int ret = 0;

#define MAX_CREATE_RETRIES 20 /* need to bound retries */
  for (int i = 0; i < MAX_CREATE_RETRIES; i++) {
    /*步骤1：初始化bucket的ver_id和quota、time等初始化信息*/
    /*步骤2：bucket_index 初始化*/
    if (zone_placement) {
      ret = svc.bi->init_index(dpp, info, info.layout.current_index);
      if (ret < 0) {
        return ret;
      }
    }

    /*步骤3：linkbucket_info信息*/
    constexpr bool exclusive = true;
    ret = put_linked_bucket_info(info, exclusive, ceph::real_time(), pep_objv, &attrs, true, dpp, y);
    if (ret == -ECANCELED) {
      ret = -EEXIST;
    }
:
    return ret;
  }

  /* this is highly unlikely */
  ldpp_dout(dpp, 0) << "ERROR: could not create bucket, continuously raced with bucket creation and removal" << dendl;
  return -ENOENT;
}
put_linked_bucket_info函数

int RGWRados::put_linked_bucket_info(RGWBucketInfo& info, bool exclusive, real_time mtime, obj_version *pep_objv,
                                     const map<string, bufferlist> *pattrs, bool create_entry_point,
                                     const DoutPrefixProvider *dpp, optional_yield y)
{
  bool create_head = !info.has_instance_obj || create_entry_point;

  /*步骤1：写bucket_instance*/
  int ret = put_bucket_instance_info(info, exclusive, mtime, pattrs, dpp, y);

  RGWBucketEntryPoint entry_point;
  entry_point.bucket = info.bucket;
  entry_point.owner = info.owner;
  entry_point.creation_time = info.creation_time;
  entry_point.linked = true;
  /*存储bucket_entrypoint实体信息*/
  ret = ctl.bucket->store_bucket_entrypoint_info(info.bucket, entry_point, y, dpp, RGWBucketCtl::Bucket::PutParams()
                                                  .set_exclusive(exclusive)
                                      .set_objv_tracker(&ot)
                                      .set_mtime(mtime));
  if (ret < 0)
    return ret;

  return 0;
}

3、SVC中的处理：bucket index的创建

int RGWSI_BucketIndex_RADOS::init_index(const DoutPrefixProvider *dpp,RGWBucketInfo& bucket_info, const rgw::bucket_index_layout_generation& idx_layout)
{
  librados::IoCtx index_pool;

  string dir_oid = dir_oid_prefix;
  int r = open_bucket_index_pool(dpp, bucket_info, &index_pool);
  if (r < 0) {
    return r;
  }

  dir_oid.append(bucket_info.bucket.bucket_id);

  map<int, string> bucket_objs;
  get_bucket_index_objects(dir_oid, idx_layout.layout.normal.num_shards, idx_layout.gen, &bucket_objs);

  return CLSRGWIssueBucketIndexInit(index_pool,
                    bucket_objs,
                    cct->_conf->rgw_bucket_index_max_aio)();
}

4、总结

至此，一个bucket创建完毕，其他的op类似于此，整体结构变化不大。下图是rgw_rados.h、rgw_sal_rados.h、rgw_service.h和svc_module***.h的相关关系，比较粗糙，仅供参考。