This blog gives you information about the minimal setup required for HANA high Availability. How to add standby host and perform a failover (simulation). How services, hosts and Volumes looks like before and after failover.
For high availability, a distributed HANA (scale out) setup is required.
The minimal setup for a scale out is 2 servers (one worker, one standby).
When an active (worker) host fails, a standby host automatically takes its place.
For that, standby host needs shared access to the database volumes.
Note: standby hosts do not contain any data and do not accept
requests or queries.
Host 1 (first node):
host role = Worker
host name = hanasl12
SID = HIA
Instance Number = 00
IP Address = 192.168.1.149
Host 2 (second node):
host role = Standby
hostname = hanadb4 (alias hanadb2)
SID = HIA
Instance number = 00
IP Address = 192.168.1.172
failover group = default
NFS is used here to shared file systems (/hana/shared, /hana/data/ and /hana/log)
export /hana from first node: /etc/exports
On second node: maintain /etc/fstab as shown below and mount the file systems.
Install SID= HIA, master node (first node) using installation media’s HDBLCM (the below screen shows services before adding standby node)
Hosts:
On master node: execute action configure_internal_network (using resident HDBLCM)
Then, on second node: run resident HDBLCM to add_hosts
Select Host role as “standby”
The below screen, shows services after adding standby node
Hosts – after adding standby node:
Volumes: (before failover) attached to active (worker) – on first node
To perform failover (simulation), I’ve killed deamon process
The below screen shows, stopped first node and now second node has Master name server (actual role) and Master Index server (actual role)
Volumes: (After failover) attached to second node
start instance on first node, it retains standby role (as actual role).
Very nice blog Satish!!
Thank you, Sridhar!!!
Hi,
Nice. This is called Auto failover testing(one of the Hana HA).
Maruthi
Thank you, Maruthi. You are right. Host auto-failover is a local fault recovery solution that can be used in addition or as an alternative measure to system replication.
Hi Sathish,
Thanks for the Blog which was help Full.
Can you please share us the basic requirements of the network side (Internet Speed for both the Linux boxes) and the Specs of servers .
Thanks,
Meghanth.S
Hi Meghanth,
My apologies for the late reply, It has been quite a while since I last checked comments here.
For production setup – 10 Gbps is required.
For minimal setup, we used 1 Gbps network.
Linux servers with 30 GB RAM and 4 CPUs (2 sockets, 2 cores per socket, 1.8 GHz) were used.
Thank you
Satish Kumar
Don’t quite understand this part :
NFS is used here to shared file systems (/hana/shared, /hana/data/ and /hana/log); export /hana from first node: /etc/exports
If the first node becomes unavailable, example power failure or OS unable to start, how is the NFS share available to the 2nd node ?
Hi Kent Peh,
valid point,
The above work was done in lab environment. My primary focus was to show Auto host failover simulation.
coming to production/non-prod setup, we have multiple options to take care of /hana/shared, /hana/data and /hana/log considering your point,
1. Non-shared SAN storage attached via Fibre Channel for HANA data (/hana/data) and log volumes (/hana/log) and shared NFS/NAS storage for HANA binaries (/hana/shared).
2. Shared storage infrastructure using NAS provides a shared everything
architecture where data volumes (/hana/data), log volumes (/hana/log) and SAP HANA binaries(/hana/shared) can be accessed via NFS.
3. Shared Storage Infrastructure using Cluster File System with local disks: A cluster file system, such as GPFS, spans the local SAP HANA server node disks, making the data and log volumes as well as the SAP HANA binaries available to the whole scaled-out landscape. In the event of a host auto-failover, the cluster file system provides a standby node with the required data set.
Thank you
Satish Kumar
Hi,
Please also help me in below design
Node 1- Own File System
Node 2 Own File System
Both Node have same hardware configuration than how to configure HA mode.
Hi, there is a way to manual switch between active and stand by with out “crash” de server or the service?
we need to test a productive environment , and I find too dangerous to use a kill process to check the stand by server. (There is a big risk to corrupt the database)
So I’m like to know if there is any command or procedure to manually aka the HANA to move the service to the stand by server. (Basically stop the active server and start into the stand by one). And then move back to the initial distribution.
Hi guys, I’m again with a question.
If I have 2 tenant running on the same active host (Tenant A and Tenant B) And I kill the service of “Tenant A”
what is the expected behavior?