Clearing hash value in GSS for a answer group

[Ralph Mendonca]

Hi,

I have a GSS rule pointing to an answer group with 2 answers. The rule uses hashed as the predictor to load balance requests between the 2 answers configured. The query for GSS comes from 2 clients. I have an issue where GSS is returning the same answer to queries from both the sources. Although both the answers are available, it seems to prefer only one answer. If I change the load balancing algorithm from hashed to Round robin, it starts returning the IP address of both the answers in round robin. The moment I change it back to Hashed, it returns only one IP address to both the sources. I suspect that the hash table is corrupted and may need to be cleared. Is there any way to do this or fix this issue? Any help on this would help.

Regards

Ralph

[sivaksiv]

Hi Ralph,

When the GSS uses the hashed balance method, elements of the client's DNS proxy IP address
and the requesting client's domain are extracted to create a unique value, referred to as
a hash value. The unique hash value is attached to and used to identify a VIP that is
chosen to serve the DNS query.

The use of hash values makes it possible to "stick" traffic from a particular requesting
client to a specific VIP, ensuring that future requests from that client are routed to the
same VIP. This type of continuity can be used to facilitate features, such as online
shopping baskets, in which client-specific data is expected to persist even when client
connectivity to a site is terminated or interrupted.

The GSS supports the following two hashed balance methods. You can apply one or both
hashed balance methods to the specified answer group.

• By Source Address—The GSS selects the answer based on a hash value created from the
source address of the request.

• By Domain Name—The GSS selects the answer based on a hash value created from the
requested domain name.

http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_services/gss4400series/v1.2/configuration/guide/Intro.html#wp1102666

For even load distribution you can use leastconns or round-robin

Regards,

Siva

[Ralph Mendonca]

Hi Siva,

Thanks for your reply but this does not answer my query. I have an issue where all users are being directed to just one answer for some reason. Ideally clients should be directed to both answers since there are 2 D-proxy clients. However, this does not seem to work. It is working for another answer group which has the same config. I would like to know if I can somehow clear the hash table so that it recalculates the hash and hopefully the load is balanced. The reason we are using hash for this config is the application requirement for a user to stick to the same server and at the same time to distribute users onto both answers even if this may not be equal since certain links in the application do not work if directed to the other server due to some information that the server caches when the user first initiates a session to the application.

Regards

Ralph

[sivaksiv]

Hi Ralph,

Was there anything changed on the DNS or D-proxy behavior?

I'm not sure if the hash table can be cleared but if its suddenly stopped working then we need to look into the tech-report and see what went wrong (if you could raise a tac case)

Regards,

Siva

[Ralph Mendonca]

Hi Siva,

Nothing was changed. It seems like I will need to log a TAC case as you said and hopefully I can get a solution to this.

Regards

Ralph