Skip to content

Commit

Permalink
Update go-time-35.md
Browse files Browse the repository at this point in the history
  • Loading branch information
Sharang Dashputre authored Dec 11, 2017
1 parent 90817aa commit 6dd7b05
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion gotime/go-time-35.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ Today on the show we have myself, Erik St. Martin, Carlisia Pinto is also here.

**Erik St. Martin:** Was this performance bottlenecks you guys were running up against, or...?

**Charity Majors:** Well, that was the one that we had first... We had a bunch of unicorns; you know, the Ruby we server... Ruby isn't exactly \[unintelligible 00:02:40.05\] so we've got this fixed pool of unicorns, and we had first one database and it was pretty okay, but by the time we had 25 replica sets behind it and MySQL and Cassandra, there was always a database breaking and you would have all of the workers filling up with in-flight requests to whatever was timing out; impossible to provision enough, to have enough in-flight workers to timeout gracefully. We tried everything to make it timeout those connections; we got into a place where it kind of mostly worked for Mongo, but it would never work for MySQL because of ActiveRecord, and it probably would have been as big of a rewrite to get out ActiveRecord, make it thread-safe, move to JRuby... Some gems were never gonna be thread-safe, and we just kind of table-flipped. We were like, "We need a threaded language."
**Charity Majors:** Well, that was the one that we had first... We had a bunch of unicorns; you know, the Ruby web server... Ruby isn't exactly thread-safe so we've got this fixed pool of unicorns, and we had first one database and it was pretty okay, but by the time we had 25 replica sets behind it and MySQL and Cassandra, there was always a database breaking and you would have all of the workers filling up with in-flight requests to whatever was timing out; impossible to provision enough, to have enough in-flight workers to timeout gracefully. We tried everything to make it timeout those connections; we got into a place where it kind of mostly worked for Mongo, but it would never work for MySQL because of ActiveRecord, and it probably would have been as big of a rewrite to get out ActiveRecord, make it thread-safe, move to JRuby... Some gems were never gonna be thread-safe, and we just kind of table-flipped. We were like, "We need a threaded language."

**Brian Ketelsen:** That is such a familiar story. I've never heard that before. \[laughter\] And I gotta be honest, I don't miss unicorns, I really don't.

Expand Down

0 comments on commit 6dd7b05

Please sign in to comment.