Rolling updates for websites hosted on Amazon's Elastic Beanstalk -
i've deployed application elastic beanstalk, , configured rolling updates feature. sounds pretty cool because keeps number of servers in farm @ times while others swapped out upgrades - seems outage-less upgrade achievable.
but reasoning isn't going work - @ point during rolling update there servers v1 on , servers v2 on. if js or css different between v1 , v2 there v1 page loaded sends request load balancer , gets v2 js/css , vice versa.
i can't think of easy way avoid this, , i'm struggling see point of rolling upgrades @ all.
am missing something? there better way achieve outage-less upgrade website? thinking set complete parallel elastic beanstalk environment v2 on it, , switch them on in 1 go - seems more time consuming.
as described, use rolling deployments , have continuous deployments on same environment, need guarantee version n compatible version n+1. "compatible" meaning can run simultaneously, can challenging in cases such different files , database schema changes.
a popular workaround blue-green deployments, deploy different environment , redirect users. swap url feature helps implement using aws elastic beanstalk.
progressive rollouts , blue-green not mutually exclusive. can use progressive rollouts small changes , blue-green bigger ones. more patterns , techniques on continuos delivery, i'd recommend check jez humble's book.
Comments
Post a Comment