Directly after we made a decision to use a managed solution that supports the Redis engine, ElastiCache quickly turned into the obvious choice. ElastiCache happy our very own two main backend specifications: scalability and reliability. The outlook of cluster balance with ElastiCache is of great interest to all of us. Before our migration, defective nodes and incorrectly balanced shards negatively influenced the availability of our backend treatments. ElastiCache for Redis with cluster-mode allowed permits us to scale horizontally with fantastic convenience.
Previously, when working with our very own self-hosted Redis structure, we’d must develop and then slashed to a totally brand-new cluster after incorporating a shard and rebalancing its slot machines. Now we initiate a scaling celebration from AWS administration system, and ElastiCache manages information replication across any extra nodes and executes shard rebalancing immediately. AWS also manages node repair (including applications patches and hardware replacement) during prepared repair happenings with limited downtime.
Eventually, we were already familiar with different merchandise from inside the AWS collection of digital products, so we understood we’re able to quickly need Amazon CloudWatch to monitor the reputation of one’s clusters.
Migration approach
Initially, we produced brand new application people to connect to the newly provisioned ElastiCache cluster. Our very own heritage self-hosted remedy used a static map of cluster topology, whereas brand-new ElastiCache-based solutions want best a major group endpoint. This new setup schema generated significantly straightforward configuration data and less upkeep across the board.
Then, we migrated production cache groups from your legacy self-hosted solution to ElastiCache by forking information writes to both groups before the brand new ElastiCache circumstances had been adequately comfortable (step 2). Here, aˆ?fork-writingaˆ? involves creating data to both the history storage in addition to new ElastiCache clusters. Nearly all of the caches need a TTL involving each admission, very in regards to our cache migrations, we usually didn’t want to perform backfills (step three) and just must fork-write both old and brand-new caches through the duration of the TTL. Fork-writes might not be essential to heated the cache incidences if the downstream source-of-truth facts sites is sufficiently provisioned to support the entire consult website traffic whilst cache are progressively inhabited. At Tinder, we normally need all of our source-of-truth storage scaled-down, and the vast majority your cache migrations call for a fork-write cache warming level. Moreover, if the TTL in the cache to-be moved is actually substantial, after that occasionally a backfill should always be accustomed facilitate 
the process.
Eventually, to possess an easy cutover while we study from your newer groups, we authenticated the fresh cluster facts by logging metrics to make sure that that information inside our brand-new caches paired that on all of our legacy nodes. Whenever we achieved a suitable threshold of congruence within replies of our legacy cache and the another one, we gradually slash more the visitors to new cache entirely (step 4). When the cutover done, we can easily reduce any incidental overprovisioning on brand new cluster.
Bottom Line
As all of our group cutovers proceeded, the frequency of node trustworthiness problems plummeted therefore practiced an age as easy as clicking various keys inside AWS Management system to measure the groups, establish new shards, and include nodes. The Redis migration freed up our procedures designers’ time and budget to a great extent and brought on remarkable improvements in tracking and automation. To learn more, read Taming ElastiCache with Auto-discovery at measure on Medium.
All of our functional and secure migration to ElastiCache provided you quick and dramatic benefits in scalability and security. We can easily never be pleased with this decision to adopt ElastiCache into all of our bunch here at Tinder.