sybase 系统进程

youwei

小弟，最近sybase系统一天之内会出现几次系统超慢的情况
看进程后发现有个进程很奇怪突然多了，许多相同的进程名称

我看了，数据库的日志
。。。。
00:00000:00000:2010/06/03 11:26:44.82 kernel  nrpacket: recv, Connection timed out
00:00000:00363:2010/06/03 14:16:37.94 kernel  Cannot read, host process disconnected:  SPBPJY 3796 spid: 363
02:00000:00537:2010/06/03 14:17:31.56 kernel  Cannot read, host process disconnected:  ZHSH 3744 spid: 537
00:00000:00491:2010/06/03 14:50:07.60 kernel  Cannot read, host process disconnected:  TONGXUN 5896 spid: 491
00:00000:01000:2010/06/03 15:27:06.54 kernel  Cannot read, host process disconnected:  SPB 3492 spid: 1000
04:00000:00027:2010/06/03 15:50:28.00 kernel  Cannot read, host process disconnected:   spid: 27
00:00000:00820:2010/06/03 15:52:48.49 kernel  Cannot read, host process disconnected:   spid: 820
00:00000:00231:2010/06/03 15:54:47.60 kernel  Cannot read, host process disconnected:  CAIZQ 3560 spid: 231
04:00000:00328:2010/06/03 16:07:07.76 kernel  Cannot read, host process disconnected:  3952 spid: 328
01:00000:00427:2010/06/03 16:27:50.90 kernel  Cannot read, host process disconnected:   spid: 427
01:00000:00270:2010/06/03 16:35:22.87 kernel  Cannot read, host process disconnected:   spid: 270
02:00000:00352:2010/06/03 16:36:40.60 kernel  Cannot read, host process disconnected:   spid: 352
06:00000:00179:2010/06/03 16:38:22.77 kernel  Cannot read, host process disconnected:  DYXINJ 3192 spid: 179
02:00000:00947:2010/06/03 16:39:10.44 kernel  Cannot read, host process disconnected:   spid: 947
01:00000:00411:2010/06/03 16:39:46.43 kernel  Cannot read, host process disconnected:  CWSHM 3304 spid: 411
。。。
好像是许多请求都被杀掉了。。
不知道是什么原因，我只有重启后又会好下。。。
数据库：sybase  12.5 for linux

wfcjz

上面的信息是你的客户端连接数据库的信息，你所说的慢是你的应用慢还是数据库里做SQL查询慢！

youwei

嗯，谢谢，小弟在系统进程里面发现有许多名称相同的进程叫
    [status]     [User]      [command]
   rece sleep      N/A       MAINTENANCE TOKE
像这样的进程有许多，一旦出现这个情况，系统在查询、登录的时候都很慢，
除了服务器的IP可以PING 通，其它什么也不行。。。
要不重启一次，要不就要等很久系统又可以登录。。

不知道 是什么原因，是数据库本身出问题了吗？
请指教，谢谢！

andkylee

回复 3# youwei


>>>>    除了服务器的IP可以PING 通，其它什么也不行。。。
不明白你这话什么意思？


下面是国内一个牛人的回复。 你可以参考一下。
  
Jason L. Froebe [Team Sybase] wrote:
>; FlyBean wrote:
>;
>;>; Hi all,
>;>;
>;>;  After a TimeSlice error, there are some processes in "runnable"
>;>; status with cmd is "MAINTENANCE TOKE".
>;>;
>;>;  I am tried of the TimeSlice error. Please refer to my privious posts.
>;>;
>;>;
>;>; Best Regards
>;>;
>;>; FlyBean
>;
>;
>; Can you post the errorlog, the select @@version, and the 'cpu grace
>; time' and 'time slice' sp_configure options?
>;
>; the sp_configure 'time slice' option should always be set to 100.  If
>; anyone tells you differently, they are wrong to do so.
>;
>; the sp_configure "cpu grace time" option should be set to about 1500.
>; This will allow for delays in hostname resolutions, etc.  The default
>; setting of 500 is far too low.
>;
>; jason

version is : ASE 12.5.2 GA + EBF12060
time slice is default.
cpu grace time is 1000.

Please refer to news://forums.sybase.com:119/41b676b7@forums-1-dub to
get more detail.

And I have found that while some connect closed not correctly may cause
this trouble. There is a bug in old OLEDB driver which the driver should
close the socket after TDS_LOGOUT without an ack token from ASE. I sure
that 1608 errors were caused by this bug.

And I will adjust the cpu grace time to 1500.

Best Regards
FlyBean

youwei

呵呵，小弟说的可以PING 通IP地址，是想说，操作系统是正常的，问题应该是在数据库上面；
上一位提出的方式，我正在试一下
sp_configure 'time slice' 这个参数的值当前是100,应该不用修改
sp_configure 'cpu grace time'  这个参数的值是500，应该是要修改为1500
我想了解一下，cpu grace time 这个参数的配置数，是根据什么来定的呢？修改了需要重启吗？
请各位多多指教，谢谢！

wfcjz

看来你的数据库需要调优，你能否把你的服务器配置说一下，然后把你的配置贴出来看一下问题出在那？

andkylee

回复 5# youwei


自己上官方网站搜索。
搜索cpu grace time 有介绍的。