don’t want to pay the IO+time of purging the data. We introduce a vttablet command line flag: -table_gc_lifecycle="hold,purge,evac,drop", that’s the default value. drop is implicit, if you don’t specify purge, no nothing. They will set -table_gc_lifecycle="drop" • some people want to keep the table around for a few days, then just go ahead and drop it: -table_gc_lifecycle="hold,drop" • some want to keep keep the table, then purge it, then immediately drop: -table_gc_lifecycle="hold,purge,drop" This is both user-customizable and simple to implement. Asking for a safe DROP Safe DROP TABLEs participate

don’t want to pay the IO+time of purging the data. We introduce a vttablet command line flag: -table_gc_lifecycle="hold,purge,evac,drop", that’s the default value. drop is implicit, if you don’t specify purge, no nothing. They will set -table_gc_lifecycle="drop" • some people want to keep the table around for a few days, then just go ahead and drop it: -table_gc_lifecycle="hold,drop" • some want to keep keep the table, then purge it, then immediately drop: -table_gc_lifecycle="hold,purge,drop" This is both user-customizable and simple to implement. Asking for a safe DROP Safe DROP TABLEs participate

don’t want to pay the IO+time of purging the data. We introduce a vttablet command line flag: -table_gc_lifecycle="hold,purge,evac,drop", that’s the default value. drop is implicit, if you don’t specify purge, no nothing. They will set -table_gc_lifecycle="drop" • some people want to keep the table around for a few days, then just go ahead and drop it: -table_gc_lifecycle="hold,drop" • some want to keep keep the table, then purge it, then immediately drop: -table_gc_lifecycle="hold,purge,drop" This is both user-customizable and simple to implement. 53 Asking for a safe DROP Safe DROP TABLEs participate

don’t want to pay the IO+time of purging the data. We introduce a vttablet command line flag: -table_gc_lifecycle="hold,purge,evac,drop", that’s the default value. drop is implicit, if you don’t specify purge, no nothing. They will set -table_gc_lifecycle="drop" • some people want to keep the table around for a few days, then just go ahead and drop it: -table_gc_lifecycle="hold,drop" • some want to keep keep the table, then purge it, then immediately drop: -table_gc_lifecycle="hold,purge,drop" This is both user-customizable and simple to implement. Asking for a safe DROP Safe DROP TABLEs participate

commits Marquee adopters 100+ contributors 700+ Slack members History ● Choosing Go ● TCP, Syscalls, GC, Scheduler 2013 2015 2016 First Kubernetes Production launch Last Sunday... Vitess Kubernetes

java/python/go Golang ? • minimalist • goroutines • channels & selects • closures • defers • generics • GC • map, slices • performace • java < go < c/c++ Vitess Features • dynamic resharding • auto-failover

the table, they see no locking issues (e.g. smaller table). Vitess supports all subsets via -table_gc_lifecycle flag to vttablet. The default is "hold,purge,evac,drop" (the complete cycle). Users may configure will always work the steps in this order: hold -> purge -> evac -> drop. For example, setting -table_gc_lifecycle "drop,hold" still first holds, then drops All subsets end with a drop, even if not explicitly those tables. The artifact or leftover tables are immediate moved to purge state. Depending on -table_gc_lifecycle, they may spend time in this state, getting purged, or immediately transitioned to the next