How to configure event logging with Dynamic Multi-Pathing 5.1 SP1

book

Article ID: 100045842

calendar_today

Updated On:

Description

Description

This document includes the following topics:

■         About DMP event logging

■         DMP event log messages

About DMP event logging

See “DMP event log messages” 

The event source daemon (vxesd) is a Veritas Dynamic Multi-Pathing (DMP) component process that receives notifications of any device-related events that are used to take appropriate actions.

DMP notifies vxesd of major events, and vxesd logs the event in a log file (/etc/vx/dmpevents.log). These events include:

■         Enabling paths or DMP nodes

■         Disabling paths or DMP nodes

■         Throttling of paths

■         I/O error analysis

■         Host Bus Adapter (HBA) and Storage Area Network (SAN) events

The log file is located in /var/adm/vx/dmpevents.log but is symbolically linked to

/etc/vx/dmpevents.log. When the file reaches 10,000 lines, the log is rotated. That is, dmpevents.log is renamed dmpevents.log.X and a new dmpevents.log file is created.

Event log messages are categorized by log level, as follows:

 

Level 1             Critical DMP log messages.

 

 

 

 

Level 2

 

Messages that relate to I/O throttling, suspected path failures, repeated path failures, and DMP node migration.

 

Level 3             Messages that relate to I/O errors, I/O error analysis, and path media errors.

 

Level 4             Messages that relate to setting or changing attributes.

 

 

Greater than level 4

 

Messages available for troubleshooting by Symantec support.

 

 

By default, only level 1 event log messages are displayed. You can change the level of detail in the event log file using the tunable dmp_log_level.

See the vxdmpadm(1M) manual page.

 

DMP event log messages

See “About DMP event logging”

Veritas Dynamic Multi-Pathing (DMP) records messages to the console and to a log file on certain events.

Table 1-1 lists the event log messages and their descriptions.

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

 

Log Level

Log Message

Meaning

Scenarios

2

Removed Path pathname belonging to Dmpnode dmpnodename

DMP deletes the path from the DMP node.

■       A Logical Unit Number (LUN) is removed from an array. All of the paths to the LUN and the DMP node are removed from the DMP database.

■       A LUN is physically migrated from one array to another. DMP detaches the paths from one enclosure and attaches them to the other.

2

I/O error occured on Path pathname belonging to Dmpnode dmpnodename

I/O did not succeed on the path

■       The path is not available physically due to errors such as an HBA controller failure, a failed LUN, or a network failure.

■       In an A/P (Active/Passive) array, the active processor is disabled.

 

 

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

(continued)

 

Log Level

Log Message

Meaning

Scenarios

3

Throttled Path pathname belonging to Dmpnode dmpnodename

The number of I/Os simultaneously processed for the DMP node has exceeded the throttle parameters.

■       If an array is slow to process concurrent I/Os, the number of I/Os may exceed the throttle level set for the DMP nodes.

DMP throttling is controlled by the following tunable parameters:

■  The queuedepth tunable parameter indicates the maximum number of I/O requests for each DMP node.

■  The iotimeout tunable parameter indicates the maximum processing time for I/Os on each path.

■       DMP throttles the path if the pending I/Os exceed either of these parameters. To tune the throttle values, use the vxdmpadm setattr command to set the queuedepth tunable parameter or the iotimeout tunable parameter.

3

Marked as idle Path pathname belonging to Dmpnode dmpnodename

An in-use path has no I/O activity for 60 seconds

■       The DMP kernel statistics daemon monitors each path’s I/O activities every 60 seconds. If the daemon does not detect I/O on a path since the last check, then the path is marked as idle.

3

Enabled Path pathname belonging to Dmpnode dmpnodename

DMP starts using the path for its operations.

■     A path is enabled manually using the

vxdmpadm enable command.

■     A new path is physically added due to the addition of an HBA or a LUN. The DMP device discovery code enables the path.

■     The DMP restore daemon enables a previously disabled path, if the inquiry and open succeeds for that path.

 

 

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

(continued)

 

Log Level

Log Message

Meaning

Scenarios

3

Disabled Path pathname belonging to Dmpnode dmpnodename

DMP stops using the path for its operations.

■     A path is disabled manually using the

vxdmpadm enable command.

■     If DMP detects an error on a path such as IO failure, inquiry or open IOCTL failure, then the path is disabled.

■     When you disable a host bus adapter (HBA) or remove a LUN, DMP disables all of the paths through the HBA or LUN.

3

SCSI error occurred on Path pathname: opcode=SCSI_operation_cCode reported transport failure (status=request_status key=sense_key, asc=additional_sense

_code, ascq=additional_ sense_code_qualifier)

DMP bypasses the SCSI strategy to issue SCSI commands to disk and the SCSI command fails with an error for the opcode

■       SCSI commands can fail if the command cannot reach the disk due to a bus error, a network error, or a SCSI adapter hardware error.

■       DMP SCSI commands like inquiry, open or failover (change from Primary to Secondary controller in an A/P or A/A-A array ) can fail with this error message.

■       There are many types of SCSI errors. Transport failure is just one of them. For complete information about the SCSI errors, refer to the following URL: http://www.isobuster.com/help.php?help=305. Uee the values of the "key", "asc" and "ascq" in the message to look up the failure in the table.

Note: For details on some typical errors, refer to the conv_scsi_tbl[] array in src/common/lib/vxvm/libvxvm/dmpapi.c

3

Marked as suspect Path pathname belonging to Dmpnode dmpnodename

DMP (vxesd daemon) has detected a change in the network configuration for the path

■     Any change in the fabric connection between the host and the array can trigger this event.

The vxesd daemon issues the VOLOP_DMP_RECONFIG_FABRIC

event to notify the vxconfigd daemon that the fabric layout has been changed which initiates device discovery.

 

 

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

(continued)

 

Log Level

Log Message

Meaning

Scenarios

3

Initiated SAN topology discovery

DMP (vxesd daemon) detects the attach or reattach of a SAN array, and initiates the device discovery

■  The vxesd daemon uses HBA-specific libraries to detect any changes in the SAN, including the addition of a LUN, removal of a LUN, or the complete removal of the SAN.

The vxesd daemon logs this message when the device discovery is started .

3

Completed SAN topology discovery

DMP (vxesd daemon) detects the attach or reattach of a SAN array, and completes the device discovery.

■  The vxesd daemon uses HBA-specific libraries to detect any changes in the SAN, including the addition of a LUN, removal of a LUN, or the complete removal of the SAN.

The vxesd daemon logs this message when the device discovery is completed.

3

Failover initiated for Dmpnode dmpnodename without/with quiescing

For A/P or A/A-A array,

DMP re-directs the I/O from one storage controller to another for the DMP node

■       The Active/Primary path on the host is manually disabled.

■       An I/O error, inquiry and open IOCTL fails on the Active/Primary path on the host.

■       The Active storage controller on the array is rebooted.

3

Reconfiguration is in progress

The vxconfigd daemon ( especially Device Discovery layer) begins the device discovery.

■  Booting the system or invoking commands like vxdisk scandisks causes vxconfigd to initiate device discovery.

■  The VOL_NOTE_RECONFIG_START event is sent to the vxesd daemon to log this event

 

 

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

(continued)

 

Log Level

Log Message

Meaning

Scenarios

3

Reconfiguration has finished

The vxconfigd daemon (

especially Device Discovery layer) completes the device discovery.

■  Booting the system or invoking commands like vxdisk scandisks causes vxconfigd to initiate device discovery.

■  When device discovery is completed, the VOL_NOTE_RECONFIG_FINISH event is sent to the vxesd daemon to log this event

3

SCSI error occurred on Path pathname: opcode=SCSI_operation_cCode reported reservation conflict (status=request_status key=sense_key, asc=additional_sense

_code, ascq=additional_ sense_code_qualifier)

DMP SCSI

command for PGR operations such as read, reserve, register, or clear failed on the path for that opcode.

■       DMP registers PGR keys on paths for Active/Active or Active/Passive cluster applications . If DMP fails to register the PGR keys, this message is logged.

■       DMP performs PGR operations for operations such as the vxdg init command, the vxdg import command, and failover ( I/O

re-direction from Primary path to Secondary path).

3

I/O retry(%u) on Path pathname belonging to Dmpnode dmpnodename

DMP error analysis detects that the path is not faulty but I/O is not successful.

DMP retries the I/O on the path

■  If I/O fails due to a device reset or busy status of disk , DMP detects that the path is not at fault. DMP receives the action DMP_PATH_OKAY from its error analysis code path. DMP then retries the I/O on the path. The number of retries is controlled by the fixed retry tunable ( dmp_retry_count) or timebound tunable (iotimeout) which can be set using the vxdmpadm setattr command.

 

 

 

Table 1-1              DMP (Veritas Dynamic Multi-Pathing) Event Log Messages

(continued)

 

Log Level

Log Message

Meaning

Scenarios

3

I/O analysis done as error on Path pathname belonging to Dmpnode dmpnodename

DMP detects an I/O error on the path. DMP analyzes the error to determine the necessary action.

■       DMP detects an I/O error if it could not send the I/O successfully to the underlying disks. This could be due to the LUN not available, network failure, or other errors.

■       Based on the I/O error, DMP analyzes the error to determine the necessary action.

For the detailed analysis error code, refer to our source code in src/common/kernel/vxdmp/dmpdefs.h

4

Unmarked as idle Path pathname belonging to Dmpnode dmpnodename

When an I/O is sent to an idle path (path that has no I/O activity), DMP removes the idle state from the path.

■     The DMP kernel statistics daemon monitors each path’s I/O activities every 60 seconds. If there are I/Os on idle paths, then DMP removes the idle state from that path.

 

Issue/Introduction

How to configure event logging with Dynamic Multi-Pathing 5.1 SP1