List of HADR new features since original HADR release in DB2 V82. Sorted by most recent ones first.
Table of Contents
-
The maximum number of standby databases available for Db2 HADR configurations is increased from three to 31.
For more information, see HADR multiple standby databases.
-
Tiebreaker policy option now available for two-node, Pacemaker-managed HADR clusters, see Setting a custom
tiebreaker policy for maintaining quorum during split-brain scenarios.
-
Node-fencing enhancement for Pacemaker-managed HADR clusters.
-
The procedure to create and configure an HADR Db2 instance on a Pacemaker-managed Linux cluster now
integrates the STONITH block device (SBD) daemon to provide a more robust fencing mechanism. This
configuration is enabled when the user adds a Qdevice to the cluster by running the db2cm cluster
manager utility. Users that are upgrading from Db2 12.1.4 and earlier can enable the same
configurations when they run the installFixPack command.
-
Simplified guide to rolling updates of HADR clusters.
-
When applying rolling updates to a Pacemaker-managed Db2 HADR environment, you no longer need to stop
or start the Pacemaker, corosync, and corosync-qdevice services on the standby and primary hosts.
With the Db2 12.1.5 release, running the installFixPack command automatically detects, stops, and
restarts these services.
-
With the release of Db2 12.1.5, the Pacemaker resource model configuration enhances when you associate a
primary VIP with an existing HADR Db2 database. The new configuration allows for earlier failover of the
VIP during force takeover operations.
-
After upgrading, you have the option of applying certain VIP configuration changes. These changes are
documented in the tech note Allow VIP failover to occur before primary reintegration. If you choose
not to update your VIP configuration, the method that the VIP follows to move nodes remains the same.
-
DNS-based traffic redirection for two-node, Pacemaker-managed Db2 HADR clusters on Amazon Web Services
(AWS). For more information, see Configuring Route 53 DNS-based traffic redirection for a
Pacemaker-managed two-node Db2 HADR cluster on AWS.
-
CPU scheduling for Pacemaker is supported in HADR. User gets this feature automatically when running db2cm
command on version 12.1.2 or later.
-
Automated HADR failover with Pacemaker is now supported on RHEL 8.2 on PPCLE and SLES 15 SP3 on PPCLE.
-
Improved primary HADR hang detection for Tivoli SA MP HADR deployments. If connections to the primary
database are unresponsive, the cluster manager will initiate a failover to the standby.
-
Multiple heartbeat rings support has been added. Users can now utilize more than one network adapter to
create multiple heartbeat rings to enhance network resiliency of the cluster. For background, benefits,
and detailed setup instructions, refer to Improving cluster network heartbeating resiliency for Db2 HADR
with Pacemaker.
-
Improved diagnostics for automated failover to HADR standby on Linux® for on-premises and
non-containerized cloud deployments with Pacemaker
-
Corosync and Pacemaker diagnostics now use a timestamp format with millisecond precision, by default,
improving traceability of sub second interactions between the Db2 engine and cluster manager.
-
There is a better solution to implement HADR in NAT environments. This method does not require the
setting of the DB2_HADR_NO_IP_CHECK register variable. This method allows HADR with multiple standby
databases to work in some NAT environments where previously is not possible via the use of
DB2_HADR_NO_IP_CHECK register variable. For more information, see HADR and Network Address Translation
(NAT) support.
-
HADR Reads on Standby new default behaviour – the availability of the standby database to perform SQL
queries is significantly enhanced through the avoidance of the replay-only window.
-
Introduced DB2_HADR_STANDBY_KEEP_UNARCHIVED_LOGS regvar
-
Controls whether standby will keep log files when the corresponding log files on primary are not
archived
-
Set to FALSE on standby, standby will delete log files when the corresponding log files on primary
are not archived
- DEFAULT: TRUE
-
HADR now respects SSL_CIPHERSPECS
-
Specifies the cipher suites that the server allows for incoming connection requests when using SSL
protocol
-
Additionally, will now also specify the cipher suites used to communicate between primary and standby
-
Changes to logarchmeth1/2 on an HADR standby will take effect upon HADR startup/re-activation as well as
during takeover.
-
In HADR Reads on Standby (ROS) environments, the availability of the standby database to perform SQL
queries is significantly enhanced through the avoidance of the replay-only window. See
DB2_HADR_ROS_AVOID_REPLAY_ONLY_WINDOW
for details.
-
A new
db2fmtlog
tool can be used to extract and display information from transaction log files. Information extracted
consists of encryption or compression status, log chain information, range of log records, and other
metadata. This information is useful in problem determination scenarios. This information can also assist
with identifying which log records can trigger 'replay only windows' in HADR Reads-on-Standby
environments. For more information, see
db2fmtlog – Format and display log file information command
-
SSL support for the transmission of transaction log data between the HADR primary and standby database
servers, previously limited to the Linux Amd64/Intel operating system, is now supported on all Unix and
Windows operating systems, excluding
Db2
pureScale environments. Setup instructions are available at
Configuring SSL for the communication between primary and standby HADR servers.
-
In HADR Standby databases with Reads On Standby (ROS) enabled, diagnostic improvements allow for easier
identification of operations which cause replay-only windows. For details on the new
db2diag.log
messaging see
DB2_HADR_REPLAY_ONLY_WINDOW_DIAGLEVEL
in
Miscellaneous variables.
For more information on the new transaction flag of
db2pd -transactions, see
db2pd - Monitor and troubleshoot Db2 database command.
-
In HADR environments, integrity checking of transaction log data during network transmission between the
primary and standby database servers is improved. When an integrity check failure is detected, seamless
retransmission of the log data is performed to auto-correct the condition, with no user intervention
required.
-
For
Db2 pureScale
users who want to upgrade from supported
Db2
Version 10.5
fix packs, or later, refer to the
FAQ technote
for supported
V10.5
fix pack levels. High availability disaster recovery (HADR) environments can now be upgraded without the
need to reinitialize the standby database after you upgrade the primary database.
-
SSL is now supported between the HADR primary and standby servers on Linux. However, this SSL support is
only present in the Linux Amd64/Intel environment, not in the Linux PPCLE or Linux Z environment. Setup
instructions are available at
Configuring SSL for the communication between primary and standby HADR servers.
-
HADR Upgrade to
Db2
11.1.1.1 from
Db2
10.5 without having to reinitialize the standby with a new backup. Further information refer to
High availability, backup, logging, resiliency, and recovery enhancements.
First available in V10.5 fp4 (Cancun release)
HADR is now supported on databases with columnar tables. Note that reads on standby still cannot query
columnar tables.
First available in V10.5
HADR is now supported on pureScale clusters. See
HADR in DB2 pureScale environments
First available in V10.5
STANDBY_SPOOL_PERCENT reports spool usage, similar to STANDBY_RECV_BUF_PERCENT.
STANDBY_RECV_BLOCKED and STANDBY_LOG_DEVICE_FULL from HADR_FLAGS are very useful warning flags.
See also
HADR monitoring
and
HADR flags monitor element
First available in V10.1
Up to 3 standbys are supported. Now you can provide HA and DR protection to your database with a single
technology. See
HADR multiple standby databases
First available in V10.1
"db2pd -hadr"
format is easier to parse.
MON_GET_HADR
table function available. New fields specifically for performance monitoring: LOG_HADR_WAIT_CUR,
LOG_HADR_WAIT_ACCUMULATED, LOG_HADR_WAIT_RECENT_AVG. See also
HADR monitoring
First available in V10.1.
Static database configuration parameters for HADR will be refreshed on database startup and on HADR
component startup. Therefore "stop HADR" followed by "start HADR" command on a primary database will pickup
new values of HADR parameters. This allows initializing and reconfiguring HADR without downtime on the
primary database. See also
HADR Config - HADR Database Configuration Parameters
First available in V10.1
Spooling allows you to specify additional space where logs can be spooled on the standby. This helps avoid
back-pressure on the primary caused by sudden spikes in logging activity. See
HADR log spooling
First available in V10.1
Delayed replay helps prevent data loss due to errant transactions. See
HADR delayed replay
First available in V95fp8 and V97fp5
In this new mode, transactions on primary database will never be slowed down or blocked. See
SUPERASYNC mode
First available in V97fp5
Reads on standby can return inline XML and LOB objects. Queries on non-inline objects will still return
error. This allows small LOB/XML objects to be returned by reads on standby.
First available in V97fp1 (Dec. 2009)
HADR standby database now accepts read-only queries. To enable this feature, set registry variable
DB2_HADR_ROS to ON on the standby instance. For details, see
Enabling reads on standby
First available in V95fp4 and V97
When database parameter BLOCKNONLOGGED is set to YES (default NO), tables with "not logged initially"
attribute or not logged LOB columns can not created, either by create table or alter table statement. See
blocknonlogged - Block creation of tables that allow non-logged activity configuration parameter
First available in V8fp17, V91fp5, and V95fp2
Db2pd -hadr option will report percentage of HADR standby buffer used. If the percentage frequently hits
100%, consider increasing buffer size. If utilization is low, consider reducing buffer size. Buffer size can
be set via registry variable DB2_HADR_BUF_SIZE. For more info, see
db2pd - Monitor and troubleshoot DB2 database command
First available in V8fp17, V91fp5, and V95fp2
Registry variable to tune socket buffer size for HADR connections. This allows tuning TCP window size for
HADR connection without impact on other TCP connections on the system. For more info, see
Miscellaneous variables
First available in V8fp16, V91fp4, and V95fp1
Registry variable for maximal primary logging wait time. Prevents primary logging from blocking because of
slow or blocked standby. For more info, see
Miscellaneous variables
First available in V95
Database configuration parameter hadr_peer_window allows takeover with no data loss when primary failed
within peer window. For more info, see
HADR peer window configuration parameter
First available in V8fp15, V91fp3, and V95.
Registry variable to bypass IP address cross check in NAT (Network Address Translation) environment. This
feature allows HADR to be deployed in NAT environment. For more info, see
HADR and Network Address Translation (NAT) support