工作中用到solr做搜索,由于没有Lucene基础,有些理论性的知识完全不能理解,不过大概还是把搜索的任务做好了,没有用到分词等功能。在使用LBHttpSolrServer的时候遇到了一个问题,就是写入的数据比预期的少一半,Solr的Wiki上这样解释LBHttpSolrServer的用法:
LBHttpSolrServer or "Load Balanced?HttpSolrServer" is just a wrapper to?CommonsHttpSolrServer. This is useful when you have multiple SolrServers and query requests need to be Load Balanced among them. It offers automatic failover when a server goes down and it detects when the server comes back up.
?
This should NOT be used for indexing in traditional master/slave architectures since updates have to be routed to the correct master. In?SolrCloud?architectures, use?CloudSolrServer?which will take advantage of this class automatically.
?
其实已经有点明白了,上面说在更新索引的时候不要使用LBHttpSolrServer了,而恰好我在更新的时候用了这个API,通过阅读源码,弄明白了LBHttpSolrServer的原理,与我预期想的不一致造成了问题的产生。我以为LBHttpSolrServer首先拿第一个可获得的节点去操作,如果当前节点挂了自动去获取下一个节点,其实LBHttpSolrServer是轮流的拿节点来操作。
?
?
public class LBHttpSolrServer extends SolrServer { private final CopyOnWriteArrayList<ServerWrapper> aliveServers = new CopyOnWriteArrayList<ServerWrapper>(); private final CopyOnWriteArrayList<ServerWrapper> zombieServers = new CopyOnWriteArrayList<ServerWrapper>(); private ScheduledExecutorService aliveCheckExecutor; private HttpClient httpClient; private final AtomicInteger counter = new AtomicInteger(-1); private ReentrantLock checkLock = new ReentrantLock(); private static final SolrQuery solrQuery = new SolrQuery("*:*"); .....
?上面摘选了LBHttpSolrServer的相关属性部门,其中值得注意的是 counter 这个属性,LBHttpSolrServer通过这个属性来记录使用LBHttpSolrServer中的哪个地址来做相关操作。通过 counter 与 存活的 server 总数取余,代码如下:
?
public NamedList<Object> request(final SolrRequest request) throws SolrServerException, IOException { int count = counter.incrementAndGet(); int attempts = 0; Exception ex; int startSize = aliveServers.size(); while (true) { int size = aliveServers.size(); if (size < 1) throw new SolrServerException("No live SolrServers available to handle this request"); ServerWrapper solrServer; try { solrServer = aliveServers.get(count % size); //这里是关键 } catch (IndexOutOfBoundsException e) { //this list changes dynamically. so it is expected to get IndexOutOfBoundsException continue; } try { return solrServer.solrServer.request(request); } catch (SolrException e) { // Server is alive but the request was malformed or invalid throw e; } catch (SolrServerException e) { if (e.getRootCause() instanceof IOException) { ex = e; moveAliveToDead(solrServer); } else { throw e; } } catch (Exception e) { throw new SolrServerException(e); } attempts++; if (attempts >= startSize) throw new SolrServerException("No live SolrServers available to handle this request", ex); } }
?其实代码中都已经详细说明了LBHttpSolrServer的相关特性,希望在用LBHttpSolrServer操作solr的时候注意好本身项目的特性,采用合适的做法。我的项目做法,不适合使用LBHttpSolrServer,原因如下:
?
我的项目没有采用SolrCloud做备份等操作,而是自己定义一个节点做所谓的master,这个负责读写索引,而在其他机器上装一个类似的环境slave,这个负责做备份,而master完全不知道有哪些节点在做它的备份,只需要在slave上执行如下命令即可同步slave与master一致:
http://slave:8984/solr/core1/replication?command=fetchindex&masterUrl=http://master:7674/solr/core1/replication
?
至于为什么采用这种方式备份,是因为SolrCloud不适合我的应用,我的应用会频繁的做大量数据的写入索引,每一次在提交时都会触发备份,导致服务器负载非常高,这里采用这种办法缓解服务器压力,只需要在slave机器上通过curl或者其他命令执行以上命令就可以完成备份操作。
?
?