StorageServer优先级的问题

Question

安装的是2.08版。
group1里有两台storage server，我想设定server1为默认源server，采用的是优先级的方法。
tracker server中设定优先级方式，然后storage server1的优先级为5，storage server2的优先级为10，这样设定有问题吗？
测试中发现，有时候group1的源server还是会轮询，连续读取的时候会取不到文件。

server的配置：
tracker1：
[2011-03-08 18:22:58] INFO - FastDFS v2.08, base_path=/home/fastdfs/tracker, connect_timeout=30s, network_timeout=60s, port=22122, bind_addr=, max_connections=256, work_threads=4, store_lookup=2, store_group=, store_server=2, store_path=0, reserved_storage_space=5120MB, download_server=1, allow_ip_count=-1, sync_log_buff_interval=10s, check_active_interval=120s, thread_stack_size=64 KB, storage_ip_changed_auto_adjust=1, storage_sync_file_max_delay=86400s, storage_sync_file_max_time=300s

tracker2：
[2011-03-08 10:24:57] INFO - FastDFS v2.08, base_path=/home/fastdfs/tracker, connect_timeout=30s, network_timeout=60s, port=22122, bind_addr=, max_connections=256, work_threads=4, store_lookup=2, store_group=, store_server=2, store_path=0, reserved_storage_space=5120MB, download_server=1, allow_ip_count=-1, sync_log_buff_interval=10s, check_active_interval=120s, thread_stack_size=64 KB, storage_ip_changed_auto_adjust=1, storage_sync_file_max_delay=86400s, storage_sync_file_max_time=300s

group1,storage1:
[2011-03-08 10:24:07] INFO - FastDFS v2.08, base_path=/home/fastdfs/storage, store_path_count=1, subdir_count_per_path=256, group_name=group1, connect_timeout=30s, network_timeout=60s, port=23000, bind_addr=, client_bind=1, max_connections=256, work_threads=4, disk_rw_separated=1, disk_reader_threads=1, disk_writer_threads=1, buff_size=256KB, heart_beat_interval=30s, stat_report_interval=60s, tracker_server_count=2, sync_wait_msec=200ms, sync_interval=0ms, sync_start_time=00:00, sync_end_time=23:59, write_mark_file_freq=500, allow_ip_count=-1, file_distribute_path_mode=0, file_distribute_rotate_count=100, fsync_after_written_bytes=0, sync_log_buff_interval=10s, sync_binlog_buff_interval=60s, sync_stat_file_interval=300s, thread_stack_size=512 KB, upload_priority=5, if_alias_prefix=, check_file_duplicate=0, FDHT group count=0, FDHT server count=0, FDHT key_namespace=, FDHT keep_alive=0, HTTP server port=8888, domain name=

group1,storage2:
[2011-03-05 10:25:55] INFO - FastDFS v2.08, base_path=/home/fastdfs/storage, store_path_count=1, subdir_count_per_path=256, group_name=group1, connect_timeout=30s, network_timeout=60s, port=23000, bind_addr=, client_bind=1, max_connections=256, work_threads=4, disk_rw_separated=1, disk_reader_threads=1, disk_writer_threads=1, buff_size=256KB, heart_beat_interval=30s, stat_report_interval=60s, tracker_server_count=2, sync_wait_msec=200ms, sync_interval=0ms, sync_start_time=00:00, sync_end_time=23:59, write_mark_file_freq=500, allow_ip_count=-1, file_distribute_path_mode=0, file_distribute_rotate_count=100, fsync_after_written_bytes=0, sync_log_buff_interval=10s, sync_binlog_buff_interval=60s, sync_stat_file_interval=300s, thread_stack_size=512 KB, upload_priority=10, if_alias_prefix=, check_file_duplicate=0, FDHT group count=0, FDHT server count=0, FDHT key_namespace=, FDHT keep_alive=0, HTTP server port=8888, domain name=

Answer 1

回复 9# happy_fish100

啊...
很有可能，我都没注意这个，谢谢，我check一下

Answer 2

回复 8# pealdoy

会不会是几台服务器的系统时间相差比较大造成的？
你确认一下系统时间是否基本一致？

Answer 3

本帖最后由 pealdoy 于 2011-03-10 09:37 编辑

回复 7# happy_fish100

2.08的呀，分了两个group，group1里有两个storage server，1和2。
部署的时候是用1当源的，但是我在测试的时候，文件有时候会存到2上去。
几率还是蛮大的，5次测试里总会碰到1次。
出错时候打出来的log：
storage server1:10.1.9.193
storage server2:10.1.9.194

上传：
upload, storage server:/10.1.9.194:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:71 - upload success. file id is: group1/M00/00/01/CgEJwk10NkuxAshVAACI-zp2apw349.txt. costs 30ms

下面4次用的是这个方法：public byte[] download_file1(String file_id)
download, storage server:/10.1.9.194:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:78 - 1, download result is: [B@1b8e059, costs 5ms
download, storage server:/10.1.9.193:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:78 - 2, download result is: null, costs 0ms
download, storage server:/10.1.9.194:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:78 - 3, download result is: [B@910040, costs 5ms
download, storage server:/10.1.9.193:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:78 - 4, download result is: null, costs 195ms

下面4次用的是这个方法：public int download_file1(String file_id, String local_filename)
download, storage server:/10.1.9.194:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:85 - 1, download result is: 0, costs 5ms
download, storage server:/10.1.9.193:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:85 - 2, download result is: 2, costs 5ms
download, storage server:/10.1.9.194:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:85 - 3, download result is: 0, costs 5ms
download, storage server:/10.1.9.193:23000, tracker server:/10.1.9.192:22122
09:35:18  INFO [fastdfs] Test:85 - 4, download result is: 2, costs 0ms

那我再检查一下服务器的配置吧

Answer 4

回复 6# happy_fish100

FastDFS server端是什么版本？
为啥我用你给出的代码做了测试，完全正常的呢？

Answer 5

回复 5# pealdoy

谢谢！
我抽空用这段代码测试一下。

Answer 6

回复 4# happy_fish100

只是一个文件上传下载的测试：

1. ClientGlobal.init(confFile);
2.         TrackerClient tracker = new TrackerClient();
3.         TrackerServer trackerServer = tracker.getConnection();
4.         StorageServer storageServer = null;
5.         StorageClient1 client = new StorageClient1(trackerServer, storageServer);
7.         NameValuePair[] metaList = new NameValuePair[1];
8.         metaList[0] = new NameValuePair("fileName", getFileName(localFile));
9.         String fileId = client.upload_file1(localFile, getExt(localFile), metaList);
10.         log.info("upload success. file id is: " + fileId);
12.         int i = 0;
13.         while (i++ < 10) {
14.                 byte[] result = client.download_file1(fileId);
15.                 log.info(i + ", download result is: " + result);
16.         }

复制代码当文件被上传到server2的时候，while里面就会轮询server2和server1，因为是刚刚上传，server1里面应该还没有文件。
还说是我每次调用之前都要从tracker server开始重新连接？

Answer 7

回复 3# pealdoy

方便贴一下相应的代码么？

Answer 8

回复 2# happy_fish100

谢谢

直接指定了storage server是什么意思？
我是直接调用的Java api中的
public byte[] download_file1(String file_id)
剩下的应该都是Java api中的代码吧？
里面会自动去找storage server的不是么？用这个newReadableStorageConnection

出现的读取错误是：
当把storage server2作为源sever的时候，
第一次取文件会从server2上下载，正常，
但是连续的第二次取会变成server1，没有文件，
第三次又会变成server2，以此轮询...

Answer 9

回复 1# pealdoy

>>测试中发现，有时候group1的源server还是会轮询
这个是程序bug，最新版本已经修复了。

>>连续读取的时候会取不到文件。
需要通过标准下载文件流程来获取文件，即先询问tracker server。
下载文件时，怀疑你直接指定了storage server。