问题
I'm brand-spanking new to the world of linux and server administration, and I'm stuck.
I have a rails app that occasionally needs to perform large data inserts, usually around 20,000 rows. The code seems to work fine in development (osx), but on the production servers (ubunto, on a linode vps), it fails every time, usually after about 1,700 insertions. The precise number varies (1655, 1697, 1756), but it's consistently in that ballpark.
I'm not seeing much that's helpful in the production.log file. just:
Connecting to database specified by database.yml
a second or so after it fails.
In the postgresql main log:
2012-10-21 23:01:28 EDT LOG: could not receive data from client: Connection reset by peer
2012-10-21 23:01:28 EDT LOG: unexpected EOF on client connection
I'm running Rails 3.2.8, ruby 1.9.3-p194, psql 1.9.4, nginx, unicorn
Really following deployment steps outlined in: http://railscasts.com/episodes/335-deploying-to-a-vps
Other notes:
a) I've tried wrapping and not wrapping the ActiveRecord insertions in a transaction. No differnece.
b) Ruby is doing a lot work to gather and organize the data before inserting to the db. This includes multiple calls to a third party web service. But I've confirmed that these communications are successful, and the data looks fine.
Any ideas? Or at least any suggestions as to where I can continue sleuthing? Thanks so much,
回答1:
The moral of the story is: "When in doubt, blame unicorns."
(Unicorn was set to timeout worker processes after 30 seconds.)
来源：https://stackoverflow.com/questions/13004664/why-is-rails-dropping-the-postgres-connection-on-large-bulk-inserts