By Valerie YJ Lee, Solutions Architect
SAP HANA is robust and has its own S/W failover feature, but if a hardware failure occurs on the infrastructure such as an H/W fault or N/W error, the data needs to be protected by an OS level HA mechanism. This document introduces how to configure SAP HANA cluster with SUSE HAE (High Availability Extension) on Alibaba Cloud.
Alibaba Cloud Enterprise-level Elastic Compute Service (ECS) instances are certified for SAP HANA. Most customers expect to configure their SAP HANA system as clustered. It works like below.
1) The two instances communicate each other via a ‘Shared Block Storage’
2) SAP request goes through HAVIP
3) The request will be processed on DB1 node
4) In abnormal status of DB1 node
i) The secondary node will take over as a primary node
ii) The other node will be restarted as a secondary node, see the changed diagram below.
- ECS instances with SUSE Linux Ent Server 12 SP2 — SUSE image for SAP application
- Shared Block Storage (Disk for sbd device)
- 2 ENIs each instance for communication channel between cluster instances
- SUSE high availability extension
Note: HAVIP and Shared Block Storage are limited for general account. By ticketing, the users will have the accessibility for these two services.
Configure SAP HANA system
Provision Alibaba Cloud infrastructure (VM level configuration)
The instance type used in this example is ‘Memory Enhanced’, which is certified on SAP HANA.
The OS is SUSE Linux Enterprise Service Pack 2. There is no public image, so a custom image is needed.
Alternatively, you can use the Marketplace image on domestic account and international account. At the time of writing, if you are using international account, the region is limited to EU Frankfurt.
Here is the network configuration like VPC, subnets, and HAVIP.
- Same VPC and zone
- HAVIP — for configuring SAP HANA HA environment, HAVIP is mandatory
- Subnets — all instances and HAVIP are in same subnet (eg. 192.168.3.0/24)
- ENI — all HANA instances must have 2 interface cards. The one is for communicate between HANA instances, and the other one is for the processing incoming request to HANA DB.
- Shared block storage — for checking hardware availability between two instances, using this shared block storage
Configuring SUSE HA Extension (OS level configuration)
Creating SBD Partition
- Initialize the SBD device with following command:
sbd -d /dev/vdg create
The SBD partition is referred to by /dev/vdg in this document. Replace it with the actual path name.
Check the SBD device with following command:
sbd -d /dev/vdg dump
Check what has been written to the device.
Follow the steps on each HANA instance
Setting up the Software Watchdog
The watchdog will protect the system against SBD failures. By default, watchdog is supported in SUSE Linux Enterprise HAE. There are some timings with SBD as STONITH device below.
Timeout (msgwait) = (Timeout (watchdog) * 2)
stonith-timeout = Timeout (msgwait) + 20%
If you need more information about the timing, please check the SBD Operation Guidelines for HAE Clusters, at this URL. https://www.suse.com/support/kb/doc.php?id=7011346.
To configure your SAP HANA Cluster correctly, there are VM level and OS level prerequisites. Make sure your account is granted for using HAVIP and Shared Block Storage. If there is no service name on the available service list, you need to open a ticket for whitelisting.
For more details about SUSE HAE overall guidance, this URL shows more:
For more details about Storage-based fencing for OS level HA configuration, this URL helps you: https://www.suse.com/documentation/sle_ha/book_sleha/data/sec_ha_storage_protect_fencing.html#pro_ha_storage_protect_sbd_daemon