HP 3000 Manuals HP 3000 Manuals
Checking and Fixing Data Sets This section explains how to check data sets for internal pointer or structural inconsistency and how to fix data sets that contain inconsistencies. To check a data set for internal or pointer inconsistency, use the CHECK SET command. To correct problems found after using CHECK SET, use the FIX SET command. You can use CHECK SET and FIX SET for either master data sets or detail data sets.
NOTE CHECK SET checks only the internal pointer consistency of the set(s) specified. It does not check the pointer consistency across sets. For example, the pointer from a chain head in a master data set to the first record on the chain in the associated detail set is not checked. To check across data sets, use the CHECK PATH command.
Checking a Master Data Set To check a master data set, use the CHECK SET command. In the following example, the master data set CUSTOMER is checked. >check set customer Check Set issued for 1 set(s). > If you want to verify that the CHECK SET command is pending, use the REVIEW SETS command. Refer to "REVIEW SETS" in chapter 10, "DBChange Plus Commands." After issuing a PERFORM COMMANDS to invoke DBAPLUS, the following output is displayed: Starting data set serial reads. Tabulating statistics. CHECK SET (MASTER) Pct Pct Longest Average Master Set Name Type Entries Capacity Full Sec Cluster Cluster ----------------------------------------------------------------------- CUSTOMER M 9 221 4 11.1 0 0.0 CHECK INFORMATION Set Name Type Search Item Message(s) ------------------------------------------------------------------------ CUSTOMER M Userlabel entry count does not match bitmap entry count (CHK 560). SALES D ACCOUNT (Path 1; linked to master CUSTOMER, path 1) Sum of chain counts in the related master set does not match bitmap entry count for this set (CHK 580). HP36386 (A.00.00) DBCHANGE (c) COPYRIGHT Hewlett-Packard Co. 1985 Analyzing the CHECK SET Statistics for a Master Data Set. CHECK SET performs a serial read followed by a chained read on the master data set specified. If you use the QUICK parameter, only a serial read is performed. The output is in two sections: CHECK SET and CHECK INFORMATION. Under the CHECK SET section of the output, the following information is reported for the above example: * The data set name. CUSTOMER is the name of the master data set. * The data set type. The following are valid data set types: M is a manual master data set. A is an automatic master data set. * Entries is the number of entries in the data set. * Capacity is the capacity of the data set. * Pct Full is the percentage of the data set capacity that is not available, rounded to the nearest whole percentage. This statistic is useful for capacity planning. The recommended percent full for a master data set is between 60 percent and 80 percent. If a master data set capacity is significantly less than 60 percent full, a serial read becomes slower. If a master data set is significantly more than 80 percent full, DBPUT intrinsics involving the master data set can slow down. If you are experiencing performance problems, consider changing the capacity of the data set. (Refer to "CHANGE SET" in chapter 10, "DBChange Plus Commands.") * Pct Sec is the percentage of secondary entries. In general, the lower the percentage, the better. A high percentage indicates that the hashing algorithm is creating many synonyms. To decrease the percentage of secondaries, increase the data set capacity to a larger number that is not a power of two. (Refer to "CHANGE SET" in chapter 10, "DBChange Plus Commands.") * Longest Cluster and Average Cluster are the longest and average number of TurboIMAGE blocks that must be read to find an open slot to place a synonym. If two records with the same hash value are added to a master data set, one of them must be placed in another available slot. Acceptable guidelines range from a cluster of 10 on a very heavily loaded transaction processing system to as much as 200 on a very lightly loaded system. A collection of historical statistics about the database can help you evaluate if the longest and average clusters are within a reasonable range. In addition, your database users can help identify clustering problems by reporting slow response time when adding records to a particular data set. Once you have determined if the range should be changed, increase or decrease the data set capacity accordingly. (Refer to "CHANGE SET" in chapter 10, "DBChange Plus Commands.") The CHECK INFORMATION section of the output identifies specific records in the data set that contain problems. If no problems are found, the following message is displayed: No problems were detected by CHECK. Fixing a Master Data Set To correct problems identified by CHECK SET, use the FIX SET command. When fixing a master data set, the following messages may occur: * Dataset DataSetName record n is duplicate manual master (FIX 170). Key value: n To correct this problem, examine the key values in the data set DataSetName and delete the key that has no entries. Be sure to print the data before deleting. * Please verify key for dataset DataSetName (FIX 170). Key value: n To correct this problem, examine the key value and the record to determine if the data is damaged. In the following example, FIX SET is used to correct the problems found in the previous CHECK SET example: >fix set customer Fix Set issued for 1 set(s). > If you want to verify that the FIX SET command is pending, use the REVIEW SETS command. Refer to "REVIEW SETS" in chapter 10, "DBChange Plus Commands." After issuing a PERFORM COMMANDS to invoke DBAPLUS, the following output is displayed: Check file ORDERSRX exists; it will be overwritten (DBA 125). Starting data set serial reads. Tabulating statistics. CHECK SET (MASTER) Pct Pct Longest Average Master Set Name Type Entries Capacity Full Sec Cluster Cluster ----------------------------------------------------------------------- CUSTOMER M 9 221 4 11.1 0 0.0 CHECK INFORMATION Set Name Type Search Item Message(s) ------------------------------------------------------------------------ CUSTOMER M Userlabel entry count does not match bitmap entry count (CHK 560). SALES D ACCOUNT (Path 1; linked to master CUSTOMER, path 1) Sum of chain counts in the related master set does not match bitmap entry count for this set (CHK 580). FIX INFORMATION ------------------------------------------------------------------------ All problems detected have been fixed. HP36386 (A.00.00) DBCHANGE (c) COPYRIGHT Hewlett-Packard Co. 1985 > The FIX SET output is similar to the CHECK SET output. The same CHECK PATH and CHECK INFORMATION reporting occurs. In addition, FIX INFORMATION reports the status of the FIX SET process. All problems found in the previous example are corrected. Checking a Detail Data Set To check a detail data set, use the CHECK SET command. In the following example, the detail data set SALES is checked: >check set sales Check Set issued for 1 set(s). > If you want to verify that the CHECK SET command is pending, use the REVIEW SETS command. Refer to "REVIEW SETS" in chapter 10, "DBChange Plus Commands." After issuing a PERFORM COMMANDS to invoke DBAPLUS, the following output is displayed: Check file ORDERSRX exists; it will be overwritten (DBA 125). Starting data set serial reads. Tabulating statistics. CHECK SET (DETAIL) Pct Highest Delete Detail Set Name Entries Capacity Full Entry Used Chain Count ----------------------------------------------------------------------- SALES 17 504 3 17 0 CHECK INFORMATION Set Name Type Search Item Message(s) ------------------------------------------------------------------------ SALES D ACCOUNT (Path 1; linked to master CUSTOMER, path 1) Path Chain inconsistencies detected (CHK 400). HP36386 (A.00.00) DBCHANGE (c) COPYRIGHT Hewlett-Packard Co. 1985 > Analyzing the CHECK SET Statistics for a Detail Data Set. CHECK SET performs a serial read followed by a chained read on the detail data set specified. If you use the QUICK parameter, only a serial read is performed. The output is in two sections: CHECK SET and CHECK INFORMATION. Under the CHECK SET section of the output, the following information is displayed for the previous example: * The data set name. SALES is the name of the detail data set. * Entries is the number of entries in the data set. * Capacity is the capacity of the data set. * Pct Full is the percentage of the data set capacity that is not available, rounded to the nearest whole percentage. This statistic is useful for capacity planning. * Highest Entry Used is the record number of the highest entry ever used (the high-water mark) in the data set. When reducing data set capacity, do not reduce it below the highest entry used; otherwise, you must repack the data set to recover the unused space in the middle of the data set. (Refer to "REPACK" in chapter 10, "DBChange Plus Commands.") * Delete Chain Count is the number of records in the delete chain. This number should be as close to 0 as possible. A high delete chain count may mean a problem with a large quantity of deletes. For example, if you have a program that performs a large number of deletes, gaps may be left in your detail data set. Subsequent record additions may disburse data randomly within the data set. As a result, serial reads may be slow. To correct a high Delete Chain Count, repack the data set. (Refer to "REPACK" in chapter 10, "DBChange Plus Commands.") The CHECK INFORMATION section of the output identifies specific records in the data set that contain problems. If no problems are found, the following message is displayed: No problems were detected by CHECK. Fixing a Detail Data Set Use the FIX SET command to correct problems identified by CHECK SET. In the following example, FIX SET is used to correct the problems found in the previous example: >fix set customer If you want to verify that the FIX SET command is pending, use the REVIEW SETS command. Refer to "REVIEW SETS" in chapter 10, "DBChange Plus Commands." After issuing a PERFORM COMMANDS to invoke DBAPLUS, the following output is displayed: Check file ORDERSRX exists; it will be overwritten (DBA 125). Starting data set serial reads. . Tabulating statistics. CHECK SET (MASTER) Pct Pct Longest Average Master Set Name Type Entries Capacity Full Sec Cluster Cluster ----------------------------------------------------------------------- CUSTOMER M 9 221 4 11.1 0 0.0 CHECK INFORMATION Set Name Type Search Item Message(s) ------------------------------------------------------------------------ CUSTOMER M Userlabel entry count does not match synonym chain count (CHK 570). SALES D ACCOUNT (Path 1; linked to master CUSTOMER, path 1) Sum of chain counts in the related master set does not match bitmap entry count for this set (CHK 580). FIX INFORMATION ------------------------------------------------------------------------ All problems detected have been fixed. PERFORM COMMANDS completed. HP36386 (A.00.00) DBCHANGE (c) COPYRIGHT Hewlett-Packard Co. 1985 > The FIX SET output is similar to the CHECK SET output. The same CHECK PATH and CHECK INFORMATION reporting occurs. In addition, the FIX INFORMATION section reports how the problem was corrected and the status of the FIX SET process. All problems found in the previous example are corrected.