Opened 7 years ago

Closed 7 years ago

#5748 closed defect (implemented)

reduce cache time on blog

Reported by: arma Owned by: phobos
Priority: Medium Milestone:
Component: Company Version:
Severity: Keywords:
Cc: Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

The blog still makes new posts unreadable for an hour.

The original reasoning for doing this was to reduce load on the blog.

Can we change it to 60 seconds? What exactly is the tradeoff here?

This issue came up again when trying to post about the latest Firefox security bug:
https://blog.torproject.org/blog/firefox-security-bug-proxy-bypass-current-tbbs

Child Tickets

Change History (5)

comment:1 Changed 7 years ago by phobos

The issue is still blog load. When it gets linked from a billion places, having the cache valid keeps the blog fast.

comment:2 Changed 7 years ago by arma

Right. Is there something magic about 3600 seconds?

Seems to me that if the goal is to almost always serve the cached version, then just about any cache timeout should be adequate. If we get 100 requests per second (major slashdotting), and the cached version lasts for 60 seconds, then once a minute we do a full request, and the other 5999 requests get the cached version.

Where does my logic break down? Is doing any real fetch at all during high load too much?

comment:3 Changed 7 years ago by phobos

Status: newaccepted

Here's the blog resource usage, https://munin.torproject.org/torproject.org/motor.torproject.org/index.html#apache

Let's put it to 5 minutes and see how it handles the load. If not, I'll keep doubling the cache time until the system is happy.

comment:4 Changed 7 years ago by phobos

As of Thu May 3 14:52:58 UTC 2012 the cache TTL is 5 minutes.

comment:5 Changed 7 years ago by phobos

Resolution: implemented
Status: acceptedclosed

we seem to have survived the onslaught of 'omg tbb is broken' press with 5m cache TTL. Closing.

Note: See TracTickets for help on using tickets.