February 2, 2017 at 7:40 am #10432
We have a cluster of A10 TH930 installed on the same site, and 3 LS internet access. We have configured:
SLB for outgoing traffic (load balancing across 3 LS internet access )
GLSB for incoming traffic : for each service (www.one.ma for example) we have created tree VIP SLB, each VIP user a LS public address, and A10 load balance the traffic across them.
We want to use the GLSB Controller to install the two devices on two sites (Primary and secondary Site = Failover Data center)
The new site (secondary site) will have a new 3 LS internet access.
For each service published by the A10 (www.one.ma for example), the internet users uses the LS1, LS2, and LS3 of primary DC (first A10), and if all theses LS (1,2 and 3) of the primary DC are down then the internet users should use the LS1, LS2 and LS3 of the secondary DC using the weited Site. Is this possible?
And if the primary site is Down, all Internet usres’request should user secondary Site Internet access.
I have a other question : We have two VIP for DNS request incoming from internet.
If the Two VIP of Primary site are down but the requested site (www.one.ma for example) is UP, what would bee the behavior of the two A10 in GLSB group. The 10 on the secondary site will response to the DNS request, but the service will be offred by the A10 in the A10 on the primary A10 ?
I need to knew the feasibility of this configuration ? and the prerequites for that.
BRMarch 1, 2017 at 12:58 pm #10682
From what I understand you want to setup “Active Standby” data centers. For this you can use Admin IP policy to always send traffic to primary site, unless it is unreachable:
Regards.March 2, 2017 at 1:12 am #10702
Regarding your 2nd question:
1. The behavior in case of the DNS VIP being unavailable is controllable via policy – you can define your zone policy to have the 2nd site respond with the primary site VIPs or you can tell the 2nd site to prefer its own VIPs – it’s configurable either way.
2. Plz consider that the condition you describe – in which the GSLB DNS VIP is unavailable (on the primary site) will most often also mean the entire site is unavailable – it’s not impossible that only the DNS VIP will be down – but it’s not really likely either.
3. In light of item 2 – it’s common for GSLB deployments to have the controllers monitor the gateways in each site – this gives a simplified manner in which to fail-over all services to DR site in case primary site is unavailable – the A10 ACOS offers the feature for monitoring the gateways.
Overall from you description your design is feasible and there are no special prerequisites except maybe diving into the GSLB policy configuration and specifically understanding how the selection algorithm works (metric-order) – check your GSLB guide for details.
Hope you find this useful.
You must be logged in to reply to this topic.