近两个月一直在测试mfs1.5.X的版本,可是在测试的过程中遇到了一些问题。
3、MFS的测试:
(1)结合昨天的测试情况,今天继续测试对于大批量文件传输的情况,并跟进资源消耗的情况,但是测试的过程中出现了问题。
使用scp指令大批量的图片时,在mfs客户端忽然发现没有了挂载的分区,重新挂载时出现:
[root@mysqldb log]# mfsmount -h 192.168.5.230 -w /usr/xxtsrc/pic/upload/
fuse: mountpoint is not empty
fuse: if you are sure this is safe, use the 'nonempty' mount option error in fuse_mount
但是查看mfs客户端的进程是依然存在的。
(2)查看master上的/va/log/message日志出现:
Jan 8 10:10:00 nginx mfsmaster[4845]: chunkservers status:
Jan 8 10:10:00 nginx mfsmaster[4845]: total: usedspace: 0 (0 GB), totalspace: 0 (0 GB), usage: 0.00%
这说明master没有发现到chunker服务器。
(3)在chunker上日志出现如下错误:
Jan 8 10:09:29 chunker1 mfschunkserver[16631]: chunk_before_io_int: file:/usr/xxtdata/5/chunk_000000000004B655_00000001.mfs - open
error (24:Too many open files)
Jan 8 10:09:29 chunker1 mfschunkserver[16631]: chunk_before_io_int: file:/usr/xxtdata/5/chunk_000000000004ADE5_00000002.mfs - open
error (24:Too many open files)
Jan 8 10:09:29 chunker1 mfschunkserver[16631]: create socket, error: Too many open files
Jan 8 10:09:43 chunker1 mfschunkserver[16631]: chunk_before_io_int: file:/usr/xxtdata/7/chunk_000000000004B047_00000002.mfs - open
error (24:Too many open files)
Jan 8 10:09:43 chunker1 mfschunkserver[16631]: chunk_before_io_int: file:/usr/xxtdata/D/chunk_000000000004B6BD_00000001.mfs - open
error (24:Too many open files)
Jan 8 10:09:44 chunker1 mfschunkserver[16631]: 3 errors occurred in 3600 seconds on folder: /usr/xxtdata/
根据提示调整了服务器的文件描述的限制,从1024加到最大,可是又发现如下问题:
Jan 8 10:35:14 chunker1 mfschunkserver[2713]: read_block_from_chunk: file:/usr/xxtdata/C/chunk_000000000004B6BC_00000002.mfs - crc
error
Jan 8 10:35:52 chunker1 mfschunkserver[2713]: read_block_from_chunk: file:/usr/xxtdata/B/chunk_000000000004B6BB_00000002.mfs - crc
error
Jan 8 10:36:31 chunker1 mfschunkserver[2713]: read_block_from_chunk: file:/usr/xxtdata/A/chunk_000000000004B6BA_00000002.mfs - crc
error
Jan 8 10:36:32 chunker1 mfschunkserver[2713]: 3 errors occurred in 3600 seconds on folder: /usr/xxtdata/
这说明有坏文件存在,致使校验没有通过,根据客户端的提示:
Jan 8 10:38:54 nginx mfsmaster[4845]: * damaged file 308729: 2009/blog/gallery_photo_s/1231/01/2009021709270130.jpg
Jan 8 10:38:57 nginx mfsmaster[4845]: damaged chunk 308920 (file: 308810 ; index: 0)
Jan 8 10:38:57 nginx mfsmaster[4845]: * damaged file 308810: 2009/blog/gallery_photo_s/1231/01/2009101608332432.jpg
Jan 8 10:38:57 nginx mfsmaster[4845]: damaged chunk 308921 (file: 308811 ; index: 0)
Jan 8 10:38:57 nginx mfsmaster[4845]: * damaged file 308811: 2009/blog/gallery_photo_s/1231/01/2008051310541895.jpg
Jan 8 10:38:57 nginx mfsmaster[4845]: damaged chunk 308922 (file: 308812 ; index: 0)
Jan 8 10:38:57 nginx mfsmaster[4845]: * damaged file 308812: 2009/blog/gallery_photo_s/1231/01/2009022508435991.jpg
Jan 8 10:38:57 nginx mfsmaster[4845]: damaged chunk 308923 (file: 308813 ; index: 0)
Jan 8 10:38:57 nginx mfsmaster[4845]: * damaged file 308813: 2009/blog/gallery_photo_s/1231/01/2009021512192253.jpg
Jan 8 10:38:57 nginx mfsmaster[4845]: damaged chunk 308924 (file: 308814 ; index: 0)
Jan 8 10:38:57 nginx mfsmaster[4845]: * damaged file 308814: 2009/blog/gallery_photo_s/1231/01/2009070309000045.jpg
将上面的损坏文件删除再次的重启之后就恢复了正常。
和mfs的SA取得联系之后,他们建议升级到1.6.X,我想问下,你那里测试的时候,有没有尽兴大批量的数据传输,或者是遇到一些问题,可以共享出来看下!