4 ms·
Sending traffic to different zones isn't the challenge, the challenge is deciding where your master data will live. In fact, this has always been one of the big
by aeden 14y ago
Sending traffic to different zones isn't the challenge, the challenge is deciding where your master data will live. In fact, this has always been one of the biggest challenges of building a fault-tolerant systems. If your master data store lives in one zone then you've got latency issues, but if it lives in multiple zones then you need to find a logical way to shard. You could also replicate across zones and then turn off writes if the zone with the master fails. You could even change masters in that case, but there's risk of data loss there.
Anyhow, sorry I don't have a simple answer - I'm not sure a simple answer exists.
- rdl 14y agoThere would be some value in failing somewhat gracefully even when your database, AZ, etc. is down. Redirect to a site-specific outage message with updates, workarounds, etc. Serving read-only also works, for some applications. The nice thing is Cloudflare does a lot of this for you for free.
- gazarsgo 14y agomulti-az in RDS does the replication and automatic failover for you, no sharding necessary.
- mh- 14y agoexcept when it doesn't -- https://forums.aws.amazon.com/thread.jspa?threadID=98376&tstart=0 https://forums.aws.amazon.com/thread.jspa?threadID=98376&...