Forum Discussion

daboochmeister's avatar
daboochmeister
Icon for Cirrus rankCirrus
Jul 27, 2017

Existing connections with Priority Group Activation - minimizing client impact, timeouts, etc.

Hi, we're using PGA for an active/standby config, with an either/or relationship between the groups (activation set to < 1 server).

 

We're finding that when the higher priority group is reactivated, existing connections to the servers in the lower priority group continue to be used for some time.

 

What are the settings that affect this? Slow ramp is involved, based on some of the answers here - persistence method? Action on service down (when a priority group is deactivated, does "action on service down" come into play?)? Other? And how does each setting come into play with PGA group activation/deactivation?

 

What settings would be required if we wanted as immediate as possible a switch back to the higher priority servers, potentially even immediate as soon as the group becomes available?

 

These questions are touched on by various questions/answers here, but no clear picture emerges (at least for me). Can point to a KB or guide that nails it all down cleanly?

 

application delivery
LTM
priority group activation

1 Reply

Sort By
  • jaikumar_f5's avatar
    jaikumar_f5
    Icon for MVP rankMVP
    Jul 29, 2017

    Did you check this article.

    Beginning in BIG-IP 11.0.0, newly-available pool members in a priority group are not counted toward the minimum-up member requirement for a pool until after the Slow Ramp Time for the pool member has expired.

    This means, your pool member A which had the highest priority would not be considered here. Slow ramp is enabled by default with 10 sec. Try disabling the slow ramp setting.

    Slow ramp is a feature where the newly available member should not be bombarded with traffic immediately but instead till the slow ramp time is reached.

    Also persistence do come into picture - Refer Article.

    When a virtual server references a priority group pool, persistence may cause traffic to persist to a node in a lower priority group even after a higher priority group is reactivated.