Record HBase Problems

Source: Internet
Author: User
At present, NoSQL products are most criticized for their stability and have to admit that HBase is far from achieving database stability (data loss, inability to read/write, DDL failure, and other serious problems ), this blog will be used to record the problems we encountered during the O & M of HBase (which will be updated continuously), hoping to help the students who use HBase. 1. A single re

At present, NoSQL products are most criticized for their stability and have to admit that HBase is far from achieving database stability (data loss, inability to read/write, DDL failure, and other serious problems ), this blog will be used to record the problems we encountered during the O & M of HBase (which will be updated continuously), hoping to help the students who use HBase. 1. A single re

At present, NoSQL products are most criticized for their stability and have to admit that HBase is far from achieving database stability (data loss, inability to read/write, DDL failure, and other serious problems ), this blog will be used to record the problems we encountered during the O & M of HBase (which will be updated continuously), hoping to help the students who use HBase.

1. When the number of region of a single regionserver is large, the write speed decreases sharply.
See http://koven2049.iteye.com/blog/1144526 for details

2. region server OOM
There were two situations that caused the region server OOM:
* The rowKey design problem is that it is always written in the same line. The version configuration is large and cannot be split. As a result, a huge file needs to be compressed during compact;
* When the application creates a table, a 3G value is set through setMaxFileSize, resulting in 6 GB space consumption for compact, thus OOM.
The reason for these two OOM is because of compact, so you need to modify compact to avoid OOM, the official 0.92 made a certain deal, specific visible: HBASE-3290.

3. master OOM
When the system has a lot of region, it is easy to cause the master OOM, see: HBASE-3906, HBase 0.90.4 or a later version of the students can ignore this problem.

4.. meta. Table hole
In hbck, the Chain of regions in table... Is broken; edges does not contain ..., This is because the startKey and endKey of the regions table do not form a closed loop, which causes some data to be read and written. In this case, the biggest problem is that you cannot fix it at will, this may cause data loss.
The reason for this problem is that offlineParentInMeta times out during split. For details, refer to: 0.90.5 or later versions to ignore this bug.
However, we are not sure to fix this bug to avoid. meta. Table hole. Therefore, we will consider a tool to safely fix this problem later.

5. Duplicate startKey/endKey in the. meta. Table
In hbck, the Chain of regions in table... Contains less elements than are listed in META; visited =, this situation is very serious. in this case, the client reads and writes may be chaotic or suspended, which may cause data loss and difficult recovery.
The cause of this phenomenon is caused by the official bug, this bug has been fixed, please refer to the HBASE-3946, use HBASE 0.90.4 or later versions of students can ignore this bug.
However, we are not sure to fix this bug to avoid repeated startKey/endKey in the. meta. Table. Therefore, we will consider a tool to safely fix this problem.

6. data may be lost when the master performs the split hlog.
For details, refer to: Workshop.

7. Writing speed decreases when reading big data.
This is because HBase's single-connection communication efficiency is low. Currently, there is no official solution to this problem. For the time being, we can only separate read/write data or make multiple connections.

8. disable table failure
If the disable table is used, the master fails and the disable fails.
The reason for the master failure is that the region in the table is not in serverAddress state, and the table is in disabling state at this time, resulting in a failure to enable, the solution can be to delete the table from the table on the zk node and then disable it. This is usually possible.
For two officially related patches, see HBASE-3892 and HBASE-4064.

9. The. meta. Table and root table are repeatedly allocated to two region servers.
For details, see http://koven2049.iteye.com/blog/1199667.

Original article address: record the HBase problems encountered. Thanks to the original author for sharing them.

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.