免费注册 查看新帖 |

Chinaunix

  平台 论坛 博客 文库
最近访问板块 发新帖
查看: 9039 | 回复: 9
打印 上一主题 下一主题

每隔10分钟就报Network connection is broken.错,不知什么原因? [复制链接]

论坛徽章:
0
跳转到指定楼层
1 [收藏(0)] [报告]
发表于 2007-02-06 19:36 |只看该作者 |倒序浏览
OS环境:solaris5.9
informix:9.4fc3

实际观察,主机没有网络故障,数据库也能正常访问,不知道什么原因会导致这样问题。

18:38:13  Checkpoint loguniq 6574, logpos 0x3ee93e8, timestamp: 231034229

18:38:13  Maximum server connections 135
18:40:07  listener-thread: err = -25582: oserr = 9: errstr = : Network connection is broken.
System error = 9.
18:48:17  Fuzzy Checkpoint Completed:  duration was 3 seconds, 831 buffers not flushed,
timestamp: 241423214.
18:48:17  Checkpoint loguniq 6574, logpos 0x49c308c, timestamp: 241423214

18:48:17  Maximum server connections 135
18:50:07  listener-thread: err = -25582: oserr = 9: errstr = : Network connection is broken.
System error = 9.
18:58:17  Fuzzy Checkpoint Completed:  duration was 1 seconds, 1360 buffers not flushed,
timestamp: 251794162.
18:58:17  Checkpoint loguniq 6574, logpos 0x51615ec, timestamp: 251794162

18:58:17  Maximum server connections 135
19:00:08  listener-thread: err = -25582: oserr = 9: errstr = : Network connection is broken.
System error = 9.

[ 本帖最后由 lumin 于 2007-2-6 19:40 编辑 ]

论坛徽章:
0
2 [报告]
发表于 2007-02-06 20:02 |只看该作者
#define EBADF            9      /* Bad file number */
查查是不是达到文件数限制

论坛徽章:
0
3 [报告]
发表于 2007-02-06 20:16 |只看该作者

感觉应该是您说的这个问题

Bad file number
===============

Generally this is a program error, not a usage error.

Contact the vendor or author of the program for an update.

Either a file descriptor refers to no open file, or a read (or
write) request is made to a file that is open only for writing
(or reading).

The symbolic name for this error is EBADF, errno=9.

但不知道导致的原因,还有解决的办法,请不吝赐教

用ulimit -n 2048 将系统最大句柄数从1024加到2048后,还是在报错

[ 本帖最后由 lumin 于 2007-2-6 20:44 编辑 ]

论坛徽章:
0
4 [报告]
发表于 2007-02-06 21:19 |只看该作者
finderr 25582
这可能表明数据库服务器已经终止或网络本身有问题。请检查网络,确认网络功能正常及服
务器正在运行中。

论坛徽章:
11
数据库技术版块每日发帖之星
日期:2016-06-25 06:20:00数据库技术版块每日发帖之星
日期:2016-06-24 06:20:00数据库技术版块每日发帖之星
日期:2016-05-03 06:20:00数据库技术版块每日发帖之星
日期:2016-04-21 06:20:00数据库技术版块每日发帖之星
日期:2016-01-23 06:20:00数据库技术版块每日发帖之星
日期:2015-12-03 06:20:00综合交流区版块每周发帖之星
日期:2015-12-02 15:03:53数据库技术版块每日发帖之星
日期:2015-10-19 06:20:00数据库技术版块每日发帖之星
日期:2015-08-20 06:20:002015年辞旧岁徽章
日期:2015-03-03 16:54:15数据库技术版块每日发帖之星
日期:2016-07-30 06:20:00
5 [报告]
发表于 2007-02-06 21:23 |只看该作者
是不是操作系统tcp的连接数不够?

论坛徽章:
0
6 [报告]
发表于 2007-02-06 21:32 |只看该作者

主机网络没有问题

TCP的连接够不够怎么看,怎么修改?

论坛徽章:
11
数据库技术版块每日发帖之星
日期:2016-06-25 06:20:00数据库技术版块每日发帖之星
日期:2016-06-24 06:20:00数据库技术版块每日发帖之星
日期:2016-05-03 06:20:00数据库技术版块每日发帖之星
日期:2016-04-21 06:20:00数据库技术版块每日发帖之星
日期:2016-01-23 06:20:00数据库技术版块每日发帖之星
日期:2015-12-03 06:20:00综合交流区版块每周发帖之星
日期:2015-12-02 15:03:53数据库技术版块每日发帖之星
日期:2015-10-19 06:20:00数据库技术版块每日发帖之星
日期:2015-08-20 06:20:002015年辞旧岁徽章
日期:2015-03-03 16:54:15数据库技术版块每日发帖之星
日期:2016-07-30 06:20:00
7 [报告]
发表于 2007-02-06 21:51 |只看该作者
你看看你的系统一开始就报这个错,还是运行一会儿后,用户数到了一定程度才报!

你看看操作系统日志里面有没有网络方面的报错!

论坛徽章:
0
8 [报告]
发表于 2007-02-06 22:00 |只看该作者

大梦

多谢!
从1月31日下午开始就一直报!每隔10分钟,非常规律。
系统日志里没有任何报错最近。
是不是将ulimit -n 2048加大了后,对已经启动的程序不生效
还是说必须改/etc/system里的rlim_fd_max值

论坛徽章:
11
数据库技术版块每日发帖之星
日期:2016-06-25 06:20:00数据库技术版块每日发帖之星
日期:2016-06-24 06:20:00数据库技术版块每日发帖之星
日期:2016-05-03 06:20:00数据库技术版块每日发帖之星
日期:2016-04-21 06:20:00数据库技术版块每日发帖之星
日期:2016-01-23 06:20:00数据库技术版块每日发帖之星
日期:2015-12-03 06:20:00综合交流区版块每周发帖之星
日期:2015-12-02 15:03:53数据库技术版块每日发帖之星
日期:2015-10-19 06:20:00数据库技术版块每日发帖之星
日期:2015-08-20 06:20:002015年辞旧岁徽章
日期:2015-03-03 16:54:15数据库技术版块每日发帖之星
日期:2016-07-30 06:20:00
9 [报告]
发表于 2007-02-06 22:49 |只看该作者
有些操作系统参数调整后是需要重启才生效!

可以先把nofiles改大试试,改了之后应该是需要重启数据库!

如果操作系统日志里面没有报错,应该和这个没什么关系!

下次重启时多启几个临听线索看看能不能解决!

若是有规律的10分钟一次,是不是你后台有一个批处理每10分钟跑一次?

对着下面检查一下看看,有没有不满足的地方!


http://publib.boulder.ibm.com/epubs/html/25120780.html

                      =================================
                            MACHINE SPECIFIC NOTES
                                    FOR
           IBM Informix Dynamic Server 9.40.FC3 PRODUCT RELEASE
                            ON SUN SOLARIS
                               DATE: 01 Dec 2003
                      =================================

PORT NUMBER: 200008 (CD ROM), 200009 (Electronic Delivery)
IRIS JOB NUMBER: 40216
OS Information : Built on Solaris 2.8, Certified on 2.9

System Requirements:
===================

1.  Solaris patch information for the IBM Informix Dynamic Server

    a. Install the recommended patch cluster for your OS version.
       The cluster is available at http://sunsolve.Sun.COM/.
       For example, if you are running Solaris 8, download and install
       8_Recommended.zip.

    b. Install the Y2000 patch cluster (also available from SunSolve)
       that applies to your OS version.

    c. Compiler-specific patches

       Due to an incompatibility between the Solaris C++ Compiler Runtime
       Version 6.0 and previous versions, the High Performance Loader
       utilities "onpladm" and "ipload", and the UNIX Blade Manager
       "blademgr" require compiler-specific patches to be installed.
       Otherwise, the following errors will be seen when invoking the
       utilities:

       ld.so.1: onpladm: fatal: relocation error: file onpladm: symbol
       __1cG__CrunSregister_exit_code6FpG_v_v_: referenced symbol not
       found

       ld.so.1: ipload: fatal: relocation error: file ipload: symbol
       __1cG__CrunSregister_exit_code6FpG_v_v_: referenced symbol not
       found

       ld.so.1: blademgr: fatal: relocation error: file blademgr:
       symbol __1cG__CrunSregister_exit_code6FpG_v_v_: referenced symbol
       not found

       The patch clusters are available at

       ftp://sunsite.unc.edu/pub/sun-info/sun-patches/

       Patches for compilers:

       Compiler version 6.0 for Solaris 8:
       108434-01
       108435-01


2.  The UNIX graphical tools provided with this release require 64-bit
    X Windows System X11R6 and Motif 2.1 or later runtime libraries.

    If these tools cannot find the location of the X and Motif shared
    libraries, the environment variable LD_LIBRARY_PATH may need to be set
    to include these locations.  For Solaris, this should be set to include
    /usr/openwin/lib/sparcv9 and /usr/dt/lib/sparcv9.

    The High Performance Loader utility "ipload" is a 32-bit binary on this
    platform.  If "ipload" cannot find the location of the 32-bit X and
    Motif shared libraries, the environment variable LD_LIBRARY_PATH may
    need to be set to include /usr/openwin/lib and /usr/dt/lib.

3.  The High Performance Loader utilities, onpload and ipload, require a
    minimum of 35 MB of memory and 70 MB of swap space in order to run.


System Configuration:
====================

1.  Kernel Parameters

    The values of the kernel parameters that were used for testing this
    product are given below.  These values might need to be tuned depending
    on the application and availability of system resources.  On Solaris,
    these entries are in /etc/system.

    set semsys:seminfo_semmap=64
    set semsys:seminfo_semmni=4096
    set semsys:seminfo_semmns=4096
    set semsys:seminfo_semmnu=4096
    set semsys:seminfo_semume=64
    set semsys:seminfo_semmsl=500
    set shmsys:shminfo_shmmax=4294967296
    set shmsys:shminfo_shmmin=100
    set shmsys:shminfo_shmmni=500
    set shmsys:shminfo_shmseg=100

    - The value of the kernel parameter "semmsl" should be set to at least
      100.  This is the maximum number of semaphores per set.  See the
      section on "Configuring Semaphore Parameters" in the IBM Informix
      Dynamic Server Migration Guide for details.

2.  Location of Shared Memory

    The ONCONFIG variable SHMBASE should be set to the following:

        SHMBASE 0x10A000000L

3.  Environment Variables

    It is recommended that the environment variable INFORMIXSTACKSIZE be
    set to 128 (the default is 32) if the application involves operations
    which would require the IBM Informix Dynamic Server to perform recursive
    database tasks (eg: cascading deletes).


4.  Audit Support

    Configuring the Operating System Audit Subsystem:
    ------------------------------------------------

    The following describes how to configure the OS audit subsystem to
    record IBM Informix Dynamic Server audit records in the OS audit log.
    In order to have OS audit support, the following steps need to be
    followed:

        a. A new audit class "ix" must be introduced to the OS audit
           subsystem as the class for IBM Informix Dynamic Server
           generated audit events.

        b. A new event "AUE_INFORMIX" must be introduced to the audit
           subsystem.  This will be an event type belonging to the "ix"
           audit class, and will be used by the IBM Informix Dynamic
           Server.

        c. The OS audit subsystem must be configured to audit events of the
           event class designated by "ix".

        d. The OS audit subsystem must be operational, and auditing enabled.

    Starting the Audit Subsystem:
    ----------------------------
    Once you have completed the configuration steps above, enable auditing
    with the new configuration.  See the operating system documentation on
    how to accomplish this.

    The OS auditing must be enabled for all users that might use the
    IBM Informix Dynamic Server, and for the event type AUE_INFORMIX,
    of the event class "ix".  See the man pages for audit, auditreduce,
    auditconfig and other related commands for details on configuring
    auditing on the system.

    The class type "ix" must be introduced by inserting a line defining the
    class in /etc/security/audit_class. The event type AUE_INFORMIX must be
    introduced to the OS with class "ix", by inserting a line defining this
    event in the file /etc/security/audit_event.

    NOTE: The entry should be added to the section for non-KERNEL events,
    and the number must be a valid number > 32767.  Pay careful attention
    to the format, even of the comments, because they are parsed
    by the auditreduce and perhaps other audit library functions at runtime.

    Setting up the System Audit Mask:
    --------------------------------
    To receive IBM Informix Dynamic Server audit data, the operating system
    audit subsystem must be configured to accept events of the type
    AUE_INFORMIX, and of class "ix".

    Line extracted from /etc/security/audit_class:
    0x00008000:ix:IBM Informix Dynamic Server Class

    Line extracted from /etc/security/audit_event:
    32768:AUE_INFORMIX:IBM Informix Dynamic Server Event:ix

    Note: If auditing is disabled at the OS level, no error will be
          generated by the IBM Informix Dynamic Server audit subsystem, as
          the system call au_write() does not return any error when auditing
          is disabled.

    Lines extracted from /etc/security/audit_control:
    dir:/var/audit
    flags:ix

    Important Files in Configuring the OS Audit Subsystem:
    -----------------------------------------------------
    /etc/security/audit_control
    /etc/security/audit_event
    /etc/security/audit_class

    Audit Log Records:
    ------------------
    To select which IBM Informix Dynamic Server events are to be audited,
    specify the audit masks using the "onaudit" utility.  Refer to the IBM
    Informix Dynamic Server Trusted Facility Manual (TFM) for further
    details.

    To extract the IBM Informix Dynamic Server audit records from the
    operating system audit trail, use the command "onshowaudit -O".  Refer
    to the IBM Informix Dynamic Server Trusted Facility Manual (TFM) for
    further details.


Feature Notes:  (alphabetized by feature)
=============

1.  High Performance Loader

    User-defined conversion routines for the High Performance Loader must be
    provided in a dynamic, shared library called ipldd09a.so.  Previously, the
    custom-code shared library file was installed in the /usr/lib directory.
    Now it is installed in the $INFORMIXDIR/lib directory. You can set the
    location of this file with the new HPL_DYNAMIC_LIB_PATH configuration
    parameter in the file pointed to by PLCONFIG environment variable. For
    more details please refer the manual, "Guide to the High-Performance Loader"

    To create a 64-bit shared library, the flags "-xarch=v9" can be used
    to convey to the compiler that a 64-bit object is to be generated.

    Onpload will use shared memory during its execution. It will allocate
    memory beyond the address for the server, leaving room for server shared
    memory expansion.

    Ipload can only connect to the IBM Informix Dynamic Server using the
    TCP/IP tli protocol.  If the IBM Informix Dynamic Server is brought
    up using shared memory ("onipcshm") protocol, ensure the following:

        a. The ONCONFIG parameter DBSERVERALIASES includes a server name
           which has "ontlitcp" as its NETTYPE.

        b. Whenever ipload is used, the environment variable INFORMIXSERVER
           should be set to the above DBSERVERALIASES.

    Refer to the IBM Informix Dynamic Server Administrator's Guide for more
    information regarding DBSERVERALIASES parameter and the INFORMIXSERVER
    environment variable.

2.  Large Memory Addressability (LMA)

    IBM Informix LMA provides the ability to support system configurations
    with greater than 4GB of RAM.  Most UNIX systems are limited to, at most,
    4GB of RAM based on the memory addressing limitations of 32-bit
    architectures.

    The values for the following ONCONFIG parameters are increased from
    32-bit to 64-bit platform by LMA support:

        - The number of BUFFERS which was limited to 768000 is now increased
          to 536870912.

        - The number of LRU queues has been increased from 128 to 512.

        - The buffer pool is allocated in a separate set of shared memory
          segments (instead of putting all the buffers in the resident and
          virtual memory segments).

        - The DS_TOTAL_MEMORY, which is the total memory available for
          decision support memory, is only limited by the amount of virtual
          memory available (2^40)/2 MB.  The sort memory comes out of the
          DS_TOTAL_MEMORY memory and hence there is no explict limit on the
          amount of sort memory.

        - The read ahead parameter RA_PAGES is only limited by BUFFERS and
          therefore can be any value less than 536870912.  The "chunk" write
          algorithm is not dependent on the amount of buffers/shared memory
          and can write as many pages in a single I/O as possible.

        - The maximum size of any shared memory segment is 4278190080 bytes.

        - The value for SHMADD which was limited to 524288 is now increased
          to 4177920. The value indicated for SHMADD parameter is in Kilo Bytes.


3.  Network Protocols

    The following protocol(s)/interface(s) are supported on this platform:

      TLI using TCP/IP
      IPC using stream pipe
      IPC using shared memory

      To use TLI with TCP/IP:
          a. NETTYPE in the ONCONFIG file and the nettype field in the
             sqlhosts file entry must contain "ontlitcp".
          b. Set the "connections" field in the ONCONFIG NETTYPE parameter
             and the kernel driver ndd parameter "tcp_conn_req_max_q" to the
             maximum number of expected concurrent connections.

             If the "tcp_conn_req_max_q" is set too low, you will see a
             client error with sqlcode = -908 and isamcode = 13.
             If the NETTYPE "connections" field is set too low, you will
             see a server listener-thread error with err = -25571.

      To use IPC (interprocess communication) with stream pipe:
          a. NETTYPE in the ONCONFIG file and the nettype field in the
             sqlhosts file entry must contain "onipcstr".

      To use IPC (interprocess communication) with shared memory:
          a. NETTYPE in the ONCONFIG file and the nettype field in the
             sqlhosts file entry must contain "onipcshm".

    Remote 32-bit applications/tools (example: ESQL/C, 4GL, ISQL) can
    only connect to the IBM Informix Dynamic Server using the TCP/IP
    protocol.

    Local 32-bit applications/tools can connect to the 64-bit server using
    either the TCP/IP tli or the IPC stream pipe protocols.

    The protocol used for supporting MaxConnect on this platform is
    "ontliimc".

4.  OnLine Backup and Restore (ON-Bar)

    Some of the functions invoked by ON-Bar are in a shared library supplied
    by third party storage management vendors.  For ON-Bar to access these
    functions, the library must be installed in one of the three ways
    noted below.

        a. Rename the library to /usr/lib/sparcv9/ibsad001.so.

        b. Create /usr/lib/sparcv9/ibsad001.so as a symbolic link to the
           library.

        c. Set the BAR_BSALIB_PATH parameter in $INFORMIXDIR/etc/$ONCONFIG
           to the full pathname of the library.

    This shared library must be compiled in 64-bit mode.

5.  OnSNMP

    The SNMP agent for OnSNMP is based on SNMP Research (SNMPR).

6.  Optical

    The name of the IBM Informix shared library for Optical on Solaris is
    iosm09a.so. This library must be located in /usr/lib/sparcv9/iosm09a.so
    (or must have a link provided in /usr/lib/sparcv9). Refer to the Guide
    to the Optical Subsystem for more information about enabling optical
    functionality.

7.  Pluggable Authentication Module (PAM)

    PAM functionality is supported on Solaris (64bit) operating system.
    The PAM is dependent on the libpam.so library supplied by the
    Operating System vendor.

    This feature introduces authentication using Pluggable Authentication
    Module (PAM) in IBM IDS 9.4.  Use of PAM can be enabled or disabled by
    the system administrator.  By default, IBM IDS will use the traditional
    authentication mechanism in order to avoid wide spread changes required
    by the users.

    The PAM service identifies the PAM module. This PAM module typically
    resides in /usr/lib/security and its parameters are listed in
    /etc/pam.conf.

8.  Simple Password Communications Support Module

    The name of the IBM Informix shared library for Simple Password CSM
    on Solaris is libixspw.so.

Features not Supported:
======================

1.  J/Foundation (Java in the Server)

    J/Foundation is not supported on this platform due to the lack of a
    64-bit JDK.

Migrating to this 64-bit Version:
================================

1.  To migrate to this 64-bit version from previous versions of the server
    (64-bit as well as 32-bit), refer to the migration process described in
    the IBM Informix Migration Guide Version 9.30.

    To revert from this 64-bit version to all versions other than 32-bit
    9.30 or 9.2x and 32-bit 9.40, refer to the IBM Informix Migration Guide
    Version 9.40.

    To revert from 64-bit 9.40 (this server) to 32-bit 9.40:
    -------------------------------------------------------
    a. Disconnect all users from this server.  The 64-bit server should
       be in online mode.

    b. Run the following command as user informix or dba:
            $INFORMIXDIR/etc/conv/update.sh -32 -d

    c. Shutdown the 64-bit server.


    d. Bring up the 32-bit 9.40 server by setting your INFORMIXDIR
       environment variable to point to the 32-bit distribution.

    To revert from 64-bit 9.40 (this server) to 32-bit 9.30 or 9.2x:
    -------------------------------------------------------
    a. Disconnect all users from this server.  The 64-bit server should be
       in online mode.


    b. Run the following command as user informix or dba:
            $INFORMIXDIR/etc/conv/update.sh -32

    c. Revert to 9.2x or 9.30 32-bit by following the steps described in the
       IBM Informix Migration Guide Version 9.40, in the section, "Migrating
       to Dynamic Server 9.40 from Dynamic Server 9.30".

2.  All UDRs and DataBlades that are built in 32-bit mode must be re-
    compiled in 64-bit mode in order to work with the 64-bit server.  If
    you have any UDRs that were developed on 32-bit, make sure that the
    proper size and alignment of the data structures are used in order for
    the UDRs to work correctly on 64-bit after re-compilation in 64-bit
    mode.

Known Defects:
=============

1.  Bug Number:         155581
    Short Description:
    CUSTOMER WOULD LIKE ONCOCKPIT PUT INTO THE 64 BIT VERSION OF THE ENGINE.

Defects Fixed in current Release:
================================


Defects Fixed in previous Release:
=================================

1.  Bug Number:      127740

    Short Description:
    IMPOSSIBLE TO CREATE A BUFFER POOL LARGER THAN 4GB (LMA) ON SOLARIS 64 BIT.

2.  Bug Number:      152157

    Short Description:
    ONBAR/ISM RESTORE MAY FAIL WITH XBSA ERROR: (BSAGETOBJECT) BACKUP OBJECT
    DOES NOT EXIST IN STORAGE MANAGER ON SOLARIS 64-BIT

3.  Bug Number:      154452

    Short Description:
    ONBAR ONBAR -V CAUSES A CORE DUMP.

4.  Bug Number:                152725

    Short Description:
    UNLOAD & LOAD TABLE USING HPL(IPLOAD) IN FIXED INTERNAL FORMAT AND THEN
    SELECT FROM THE TABLE RETURNS 0 FOR AN INTEGER COLUMN

5.  Bug Number:         155563

    Short Description:
    SELECT STATEMENTS CRASHED THE ENGINE IN GET_TMPVALMEMORY FUNCTION .

6.  Bug Number:         154456

    Short Description:
    -9791 DATA DICTIONARY HASH IS EXCEEDED LDNAME GETS ASSERT FAILURE
    MTEX.C LINE 368 WHEN INSERT/UPDATE TABLE WITH TRIGGER CALLING STORED
    PROCEDURE.

[ 本帖最后由 大梦 于 2007-2-6 23:04 编辑 ]

论坛徽章:
0
10 [报告]
发表于 2007-03-02 14:57 |只看该作者
是不是buffer大了?
您需要登录后才可以回帖 登录 | 注册

本版积分规则 发表回复

  

北京盛拓优讯信息技术有限公司. 版权所有 京ICP备16024965号-6 北京市公安局海淀分局网监中心备案编号:11010802020122 niuxiaotong@pcpop.com 17352615567
未成年举报专区
中国互联网协会会员  联系我们:huangweiwei@itpub.net
感谢所有关心和支持过ChinaUnix的朋友们 转载本站内容请注明原作者名及出处

清除 Cookies - ChinaUnix - Archiver - WAP - TOP