18.1.1 Actions to be taken by the HiRDB administrator when an error occurs

This section explains the actions to be taken by the HiRDB administrator when an error occurs.

Organization of this subsection
(1) Actions to be taken by the HiRDB administrator
(2) Executing the pdgeter command
(3) Deleting unneeded troubleshooting information
(4) Reducing the amount of troubleshooting information that is output

(1) Actions to be taken by the HiRDB administrator

When an error occurs, the HiRDB administrator must take the following actions:

  1. Check the output messages and the manual HiRDB Version 8 Messages to determine the causes of the error.
  2. Use the pdgeter command to make a backup copy of the files under $PDDIR/spool and $PDDIR/tmp to which troubleshooting information has been output
Remarks

(2) Executing the pdgeter command

When the pdgeter command is executed, a backup copy of the troubleshooting information is created under a directory in the server machine where the pdgeter command was entered. Any directory can be specified in an option of the pdgeter command. The backup copy can also be created at any device, not just in the server machine used to enter the pdgeter command (not supported in the AIX 5L version). The following is the procedure for executing the pdgeter command:

Procedure
  1. Check that the server machine where the pdgeter command is entered has sufficient disk space to store the troubleshooting information.
  2. Create a directory* for storing the troubleshooting information at the server machine where the pdgeter command is entered. This step is not necessary if the troubleshooting information is to be output to a device.
  3. If troubleshooting information for a machine other than the server machine where the pdgeter command is entered is to be output to a device, a work directory is required. In such a case, create a work directory* in the server machine where the pdgeter command is entered.
  4. Copy the pdinit control statement to $PDDIR/conf, using INITCONT as the file name. If all troubleshooting information is to be obtained (by specifying the -a option), copy the information under the file name INITCONT into $PDDIR/conf at the server machine where the pdgeter command is entered.
  5. Copy the pdmod control statement into $PDDIR/conf. In this case, any file name can be used.
  6. If shared memory is to be dumped (by omitting the -m option from the pdgeter command), check that there is a directory for storing a shared memory dump ($PDDIR/spool/pdshmdump) at the server machine subject to collection of troubleshooting information (the server machine with the host name specified in the -x option of the pdgeter command). If there is no such directory, create it.
  7. Execute the pdgeter command. Specify the directories provided in steps 2 and 3 in the pdgeter command's options.
  8. When the troubleshooting information is no longer needed, delete it with the pdcspool command (retaining it may result in a shortage of disk space).
* If this directory already exists, check whether or not it contains any of the directories or files listed below; if it contains these directories or files, the pdgeter command will result in an error.
  • PDDIR
  • lib
  • usr
  • HiRDB

(3) Deleting unneeded troubleshooting information

When a server process or client is terminated forcibly, HiRDB outputs troubleshooting information to the $PDDIR/spool directory. In addition, whenever the Ctrl + C keys are pressed to terminate a command or a utility while it is executing, the command or utility outputs a temporary work file to the $PDDIR/tmp directory, where it remains resident. If these troubleshooting and temporary work files are left on the disk, they may stress the capacity of the disk on which the HiRDB directory resides. A shortage of free space on the disk containing the HiRDB directory can cause HiRDB to terminate abnormally. To avoid such a problem, HiRDB deletes the following files periodically:

Normally, HiRDB deletes these files every 24 hours. You can change this deletion interval in the pd_spool_cleanup_interval operand. You can also specify that only files that were output a specified number of days previous to the current date are to be deleted; you make this specification in the pd_spool_cleanup_interval_level operand.

You can also use the following methods to all delete troubleshooting information (files in the $PDDIR/spool directory):

(4) Reducing the amount of troubleshooting information that is output

You can specify the following operands to reduce the amount of troubleshooting information that is output. Specify these operands as needed.