Varnish trouble log

This page contains a chronological list of noteworthy problems in Varnish.

2010-06-28 GCC issue on Solaris

GCC 3.4.3 as shipped by Sun seems to run into some pretty esoteric #include file trouble.

The solution is to get hold of GCC 3.4.6 or later.

See ticket #711 for details.

2010-06-18 Object header leak in 2.1.2 and trunk

There seem to be a leak of object header objects in 2.1.2 (probably 2.1 as well). If you experience increased CPU usage and slowdowns try switching adding '-hclassic,$BIGPRIME' to the command line where $BIGPRIME is a prime number to indicate the number of buckets you need (try with ~ 10% of the number of objects you expect to store).

2010-05-04 2.1.1+HEAD bug may append junk to objects

If an object was delivered from the backend using chunked encoding, and was larger than the storage segment size (default: 128k) this bug may bite.

The effect of the bug is that up to storage segment worth of junk may be appended to the transmitted object.

This is mostly harmless, because the Content-Length header will make the browser do the right thing, but certain load-balancers will go cross-eyed and act really weird at the TCP level, spewing interesting RST packets to the client.

To fix this for good, you can apply the one-line patch in r4758.

As a workaround, you can increase the "fetch_chunksize" parameter to be bigger than any object your backend may deliver with chunked encoding. Be aware that this will affect storage usage negatively.

Telling your backend to not use chunked encoding (if possible) is also a water-tight workaround.

This bug is not in 2.1.0

2010-04-15 HEAD requests converted to GET for pipe+pass in 2.1.0

2009-12-02 ESI more or less broken in 2.0.5

  • See #585
  • Releases affected: 2.0.5

2008-10-20 Core dump in ESI processing code

  • First affected: r2069 (when ESI was introduced)
  • Fixed in: r3322
  • Releases affected: 2.0, 2.0.1

ESI:included objects gets mishandled on the waiting list if another threads holds the included object busy while contacting the backend.

2008-10-17 Core dump on invalid HTTP protocol input

  • First affected: r3290
  • Fixed in: r3315
  • Releases affected: 2.0

Certain invalid HTTP requests can crash Varnish with little or no usable diagnotics.

2008-10-16 ACL compilation error for CIDR masks

  • First affected: r2295
  • Fixed in: r3308
  • Releases affected: all 2.x until and including 2.0

ACL rules with a specified mask different from /8, /16, /24 ... /56 do not work correctly.

Such rules will both miss intended target IP numbers and hit unintended IP numbers.

2008-08-19 Wrong headers can be sent to backend

  • Ticket: #292
  • Fixed in: r3110
  • Releases affected: All 1.x, 2.0-tp1, 2.0-tp2

If the client sends a Connection: header and the VCL code deletes one or more HTTP headers, the wrong set of headers may be sent to the backend.

The typical result is that the next header after the one listed in Connection: is deleted, but depending on the exact headers, more complex behaviour can arise, for instance if the Host: header is deleted this way.