Ceph

ceph rgw：object tagging实现

2017-12-05  本文已影响47人  chnmagnus

object tagging也是L版本新增加的特性，在L版本的release notes中描述如下：

S3 Object Tagging API has been added; while APIs are supported for GET/PUT/DELETE object tags and in PUT object API, there is no support for tags on Policies & Lifecycle yet

主要关注点是

  1. PUT/GET/DELETE Object Tags的实现和使用情况
  2. 与lifecycle或policy结合的功能的实现情况

实现

rgw对tagging的实现比较简单，一个rgw object的tags就存储在这个rgw object对应的rados的header对象的xattr中。使用时也是直接从对应对象的xattr查询。相关代码片段见附录。

关注点1：PUT/GET/DELETE Object Tags操作

创建用于测试的bucket和object：testbukcet/testobj。
用curl模拟PUT/GET/DELETE Object Tags的http请求。

PUT Object Tags

将tagging信息以xml格式存在文件中，使用PUT方法设置object tags。
文档：http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectPUTtagging.html

 #!/bin/bash
 
access_key="0555b35654ad1656d804"
secret_key="h7GhxuBLTrlhVUyxSPUKUV8r/2EI4ngqJxD7iBdBYLhwluN30JaT3Q=="
date=$(date -R -u)
string_to_sign="PUT\n\n\n${date}\n/testbucket/testobj?tagging"
signature=$(echo -en ${string_to_sign} | openssl sha1 -hmac ${secret_key} -binary | base64)
content_length="126"

curl "http://127.0.0.1:8000/testbucket/testobj?tagging"             \
    -H "Content-Length: ${content_length}"                          \
    -H "Date: ${date}"                                              \
    -H "User-Agent: cmh-test"                                       \
    -H "Authorization: AWS ${access_key}:${signature}"              \
    -H "Expect:"                                                    \
    -X PUT -T ./testtagging.xml -v

GET Object Tags

文档：http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectGETtagging.html

#!/bin/bash

access_key="0555b35654ad1656d804"
secret_key="h7GhxuBLTrlhVUyxSPUKUV8r/2EI4ngqJxD7iBdBYLhwluN30JaT3Q=="
date=$(date -R -u)
string_to_sign="GET\n\n\n${date}\n/testbucket/testobj?tagging"
signature=$(echo -en ${string_to_sign} | openssl sha1 -hmac ${secret_key} -binary | base64)

curl "http://127.0.0.1:8000/testbucket/testobj?tagging"             \
    -H "Date: ${date}"                                              \
    -H "User-Agent: cmh-test"                                       \
    -H "Authorization: AWS ${access_key}:${signature}"              \
    -H "Expect:"                                                    \
    -X GET -v

DELETE Object Tags

文档：http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectDELETEtagging.html

#!/bin/bash

access_key="0555b35654ad1656d804"
secret_key="h7GhxuBLTrlhVUyxSPUKUV8r/2EI4ngqJxD7iBdBYLhwluN30JaT3Q=="
date=$(date -R -u)
string_to_sign="DELETE\n\n\n${date}\n/testbucket/testobj?tagging"
signature=$(echo -en ${string_to_sign} | openssl sha1 -hmac ${secret_key} -binary | base64)

curl "http://127.0.0.1:8000/testbucket/testobj?tagging"             \
    -H "Date: ${date}"                                              \
    -H "User-Agent: cmh-test"                                       \
    -H "Authorization: AWS ${access_key}:${signature}"              \
    -H "Expect:"                                                    \
    -X DELETE -v

根据前面所说，tagging的内容是存储在testobj对应的rados 对象的xattr中的，在增删tagging前后，可以使用下面的命令看查看testobj的xattr增改。

$ ./bin/rados -p default.rgw.buckets.data listxattr 38d08ed7-3883-49de-ab89-0dea7c8c960f.4162.2_testobj

关注点2：lifecycle根据tag进行对象淘汰

这部分在L版本的release notes的描述中，是不被支持的。但在master版本的代码中是有写的，具体见附录最后“lifecycle机制中的tagging应用”。

附录：代码片段

RGWPutObjTags::execute()

设置obj tagging的代码，可以看到，代码将tag信息以RGW_ATTR_TAGS作为key，存储在对应object的xattr中。

void RGWPutObjTags::execute(){
  op_ret = get_params();
  if (op_ret < 0)
    return;

  if (s->object.empty()){
    op_ret= -EINVAL; // we only support tagging on existing objects
    return;
  }

  rgw_obj obj;
  obj = rgw_obj(s->bucket, s->object);
  store->set_atomic(s->obj_ctx, obj);
  op_ret = modify_obj_attr(store, s, obj, RGW_ATTR_TAGS, tags_bl);
  if (op_ret == -ECANCELED){
    op_ret = -ERR_TAG_CONFLICT;
  }
}

RGWGetObjTags::execute()

获取tag的代码，可以看到，代码获取了对应object的xattr，然后通过RGW_ATTR_TAGS作为key读取。

void RGWGetObjTags::execute(){
  rgw_obj obj;
  map<string,bufferlist> attrs;

  obj = rgw_obj(s->bucket, s->object);

  store->set_atomic(s->obj_ctx, obj);

  op_ret = get_obj_attrs(store, s, obj, attrs);
  if (op_ret < 0) {
    ldout(s->cct, 0) << "ERROR: failed to get obj attrs, obj=" << obj
             << " ret=" << op_ret << dendl;
    return;
  }

  auto tags = attrs.find(RGW_ATTR_TAGS);
  if(tags != attrs.end()){
    has_tags = true;
    tags_bl.append(tags->second);
  }
  send_response_data(tags_bl);
}

RGWDeleteObjTags::execute()

void RGWDeleteObjTags::execute(){
  if (s->object.empty())
    return;

  rgw_obj obj;
  obj = rgw_obj(s->bucket, s->object);
  store->set_atomic(s->obj_ctx, obj);
  map <string, bufferlist> attrs;
  map <string, bufferlist> rmattr;
  bufferlist bl;
  rmattr[RGW_ATTR_TAGS] = bl;
  op_ret = store->set_attrs(s->obj_ctx, s->bucket_info, obj, attrs, &rmattr);
}

lifecycle机制中的tagging应用

在worker线程遍历lifecycle对应的bucket下的objects时，其中有一步就是查询该对象xattr中的tags，与lifecycle规则中的tags进行对比。

bucket_lc_process函数代码片段：

if (prefix_iter->second.obj_tags != boost::none) {
  bufferlist tags_bl;
  // 从object的xattr中获得tags
  int ret = read_obj_tags(store, bucket_info, obj, rctx, tags_bl);
  // read_obj_tags的代码实现，可以看到，它做的就是获取对象的xattr，并使用RGW_ATTR_TAGS作为key获得tags对应的entry。
  /*  
      static int read_obj_tags(RGWRados *store, RGWBucketInfo& bucket_info, rgw_obj& obj, RGWObjectCtx& ctx, bufferlist& tags_bl)
      {
        RGWRados::Object op_target(store, bucket_info, ctx, obj);
        RGWRados::Object::Read read_op(&op_target);
        return read_op.get_attr(RGW_ATTR_TAGS, tags_bl);
      }
  */
  RGWObjTags dest_obj_tags;
  try {
    auto iter = tags_bl.begin();
    // 将bufferlist中的tags信息decode到dest_obj_tags中
    dest_obj_tags.decode(iter);
  } catch (buffer::error& err) {
     ldout(cct,0) << "ERROR: caught buffer::error, couldn't decode TagSet" << dendl;
    return -EIO;
  }
  // 判断当前遍历的object的tags和lifecycle中指定的tags是否相同，如果不相同则跳过该object，否则进行后续的其他验证和删除操作
  if (!includes(dest_obj_tags.get_tags().begin(),
                dest_obj_tags.get_tags().end(),
                prefix_iter->second.obj_tags->get_tags().begin(),
                prefix_iter->second.obj_tags->get_tags().end())){
    ldout(cct, 20) << __func__ << "() skipping obj " << key << " as tags do not match" << dendl;
    continue;
  }
}
上一篇下一篇

猜你喜欢

热点阅读