ceph中用到的压缩引擎
ceph可以默认支持snappy/zlib/zstd,如果要支持lz4的话,必须打开宏HAVE_LZ4
set(compressor_srcs 
  Compressor.cc)
add_library(compressor_objs OBJECT ${compressor_srcs})

## compressor plugins

set(compressor_plugin_dir ${CMAKE_INSTALL_PKGLIBDIR}/compressor)

add_subdirectory(snappy)
add_subdirectory(zlib)
add_subdirectory(zstd)

if (HAVE_LZ4)
  add_subdirectory(lz4)
endif()

从源码压缩又分为同步和异步压缩两种,其中异步压缩也是调用同步压缩的接口,只是不去
等待结果就返回了
这里我们以异步压缩为例
AsyncCompressor::AsyncCompressor(CephContext *c):
  #这里调用同步接口conpressor新建一个压缩engine
  compressor(Compressor::create(c, c->_conf->async_compressor_type)), cct(c),
  #新建一个thread pool 用来处理压缩和解压
  compress_tp(cct, "AsyncCompressor::compressor_tp", "tp_async_compr", cct->_conf->async_compressor_threads, "async_compressor_threads"),
  #定义锁
  job_lock("AsyncCompressor::job_lock"),
  #定义一个workqueue,要与要压缩和解压的工作
  compress_wq(this, c->_conf->async_compressor_thread_timeout, c->_conf->async_compressor_thread_suicide_timeout, &compress_tp) {
}
在其init 函数中启动压缩用的thread pool
void AsyncCompressor::init()
{
  ldout(cct, 10) << __func__ << dendl;
  compress_tp.start();
}
调用如下接口开始压缩
uint64_t AsyncCompressor::async_compress(bufferlist &data)
{
  uint64_t id = ++job_id;
  pair<unordered_map<uint64_t, Job>::iterator, bool> it;
  {
    Mutex::Locker l(job_lock);
	#每个压缩的数据对应一个id
    it = jobs.insert(make_pair(id, Job(id, true)));
    it.first->second.data = data;
  }
  #可以看到这里讲要压缩的数据组成job 放到wq中就算返回了,并没有等待压缩完成在返回,这就是同步和异步压缩的最大的区别
  compress_wq.queue(&it.first->second);
  ldout(cct, 10) << __func__ << " insert async compress job id=" << id << dendl;
  return id;
}

  return id;
}

要获取压缩后的数据可以调用下面的接口来得到

int AsyncCompressor::get_compress_data(uint64_t compress_id, bufferlist &data, bool blocking, bool *finished)
{
  assert(finished);
  Mutex::Locker l(job_lock);
  #根据开始压缩时候的id来查找压缩后的数据
  unordered_map<uint64_t, Job>::iterator it = jobs.find(compress_id);
  #如果没有找到id则返回error
  if (it == jobs.end() || !it->second.is_compress) {
    ldout(cct, 10) << __func__ << " missing to get compress job id=" << compress_id << dendl;
    return -ENOENT;
  }

 retry:
  auto status = it->second.status.load();
  #压缩已经完成,则返回数据,并删除这个job
  if (status == status_t::DONE) {
    ldout(cct, 20) << __func__ << " successfully getting compressed data, job id=" << compress_id << dendl;
    *finished = true;
    data.swap(it->second.data);
    jobs.erase(it);
	#压缩 失败,也删除这个job,返回error
  } else if (status == status_t::ERROR) {
    ldout(cct, 20) << __func__ << " compressed data failed, job id=" << compress_id << dendl;
    jobs.erase(it);
    return -EIO;
  } else if (blocking) {

    auto expected = status_t::WAIT;
    if (it->second.status.compare_exchange_strong(expected, status_t::DONE)) {
      ldout(cct, 10) << __func__ << " compress job id=" << compress_id << " hasn't finished, abort!"<< dendl;
	    #正在被block中,如果没有abort,则调用同步压缩接口等待压缩完成,这个时候异步压缩退化为同步压缩
      if (compressor->compress(it->second.data, data)) {
        ldout(cct, 1) << __func__ << " compress job id=" << compress_id << " failed!"<< dendl;
        it->second.status = status_t::ERROR;
        return -EIO;
      }
      *finished = true;
    } else {
	#等待1s在检查数据是否压缩完成
      job_lock.Unlock();
      usleep(1000);
      job_lock.Lock();
      goto retry;
    }
  } else {
    ldout(cct, 10) << __func__ << " compress job id=" << compress_id << " hasn't finished."<< dendl;
    *finished = false;
  }
  return 0;
}

阅读更多
版权声明:本文为博主原创文章,未经博主允许不得转载。 https://blog.csdn.net/tiantao2012/article/details/80346813
个人分类: ceph
想对作者说点什么? 我来说一句

ceph数据读写原理

2017年08月24日 713KB 下载

ceph常见运维指令

2014年09月19日 23KB 下载

没有更多推荐了,返回首页

不良信息举报

ceph中用到的压缩引擎

最多只允许输入30个字

加入CSDN,享受更精准的内容推荐,与500万程序员共同成长!
关闭
关闭