Shulou(Shulou.com)06/03 Report--

I recently encountered a file permission problem in solr, which caused the index to crash. Colleagues also encountered a similar phenomenon in FS. After many experiences, it has been found that hadoop currently has synchronization problems with directory rights management.

Normally, if you start with a user, the directory permissions become that user. As for the group in which the user belongs, it can be ignored directly and does not see the use. But sometimes this happens. Obviously run the program is a super user, that is, the default use of hdfs users (if there is a default), but can access solr, permissions directory, and has been running steadily.

But one day, you may suddenly want to change some configuration. Restarted some apps. Maybe all of a sudden, its authority changed. Unless you start with the correct user and correct the problems caused by permissions, you will not be able to access it. I have experienced it twice, once with cloudear solr (solrcloud). The other time is an extension application of hdfs, encountered in the file system. All of them are available at the beginning, but suddenly can't be used after directory cleaning, or configuration changes, and restarting some applications. Looks like a delayed landmine. A delayed explosion of thunder.

Think carefully about the whole mechanism of hadoop, the mechanism of hdfs system, and the management mechanism of linux file system. Finally, I think that may be the reason.

The hdfs file system, mainly nameserver, does not take effect immediately when managing permissions, unless you use chmod to change permissions. It has a trigger condition for updating data. Prior to this, all files were accessible across permissions. And it allows it to do so. This is because it avoids excessive synchronization and improves the efficiency of file system management. It also reduces a series of problems caused by synchronization. Strictly speaking, its permission synchronization is triggered manually. It will not be updated unless you make it clear that I need to limit changes only.

In this way, all started processes, all files that are already active in namenode, can be accessed across permissions. But in occasional cases, it triggers permission synchronization, causing subsequent applications to suddenly become inaccessible, showing that the directory has no permissions.

At this point, you need to manually update all directory permissions. It is recommended to restart all service nodes and all related applications at the same time.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.