Achieving a high hitrate

Now that Varnish is up and running you can access your web application through Varnish. Unless your application is specifically written to work behind a web accelerator you'll probably need to do some changes to either the configuration or the application in order to get a high hitrate in Varnish.

Varnish will not cache your data unless it's absolutely sure it is safe to do so. So, for you to understand how Varnish decides if and how to cache a page, we'll guide you through a couple of tools that you should find useful to understand what is happening in your Varnish setup.

Note that you need a tool to see the HTTP headers that fly between Varnish and the backend. On the Varnish server, the easiest way to do this is to use varnishlog and varnishtop but sometimes a client-side tool makes sense. Here are the ones we commonly use.

Tool: varnishtop

You can use varnishtop to identify what URLs are hitting the backend the most. varnishtop -i BereqURL is an essential command, showing you the top requests Varnish is sending to the backend. You can see some other examples of varnishtop usage in Statistics.

Tool: varnishlog

When you have identified an URL which is frequently sent to the backend you can use varnishlog to have a look at the request. varnishlog -q 'ReqURL ~ "^/foo/bar"' will show you the requests coming from the client matching /foo/bar.

For more information on how varnishlog works please see Logging in Varnish or then man page.

For extended diagnostics headers, see https://www.varnish-cache.org/trac/wiki/VCLExampleHitMissHeader

Tool: lwp-request

lwp-request is tool that is a part of The World-Wide Web library for Perl. It's a couple of really basic programs that can execute an HTTP request and show you the result. We mostly use the two programs, GET and HEAD.

vg.no was the first site to use Varnish and the people running Varnish there are quite clueful. So it's interesting to look at their HTTP Headers. Let's send a GET request for their home page:

$ GET -H 'Host: www.vg.no' -Used http://vg.no/
GET http://vg.no/
Host: www.vg.no
User-Agent: lwp-request/5.834 libwww-perl/5.834

200 OK
Cache-Control: must-revalidate
Refresh: 600
Title: VG Nett - Forsiden - VG Nett
X-Age: 463
X-Cache: HIT
X-Rick-Would-Never: Let you down
X-VG-Jobb: http://www.finn.no/finn/job/fulltime/result?keyword=vg+multimedia Merk:HeaderNinja
X-VG-Korken: http://www.youtube.com/watch?v=Fcj8CnD5188
X-VG-WebCache: joanie
X-VG-WebServer: leon

OK. Lets look at what GET does. GET usually sends off HTTP 0.9 requests, which lack the 'Host' header. So we add a 'Host' header with the '-H' option. '-U' print request headers, '-s' prints response status, '-e' prints response headers and '-d' discards the actual content. We don't really care about the content, only the headers.

As you can see, VG adds quite a bit of information in their headers. Some of the headers, like the 'X-Rick-Would-Never' are specific to vg.no and their somewhat odd sense of humour. Others, like the 'X-VG-Webcache' are for debugging purposes.

So, to check whether a site sets cookies for a specific URL, just do:

GET -Used http://example.com/ |grep ^Set-Cookie

Tool: Live HTTP Headers

There is also a plugin for Firefox called Live HTTP Headers. This plugin can show you what headers are being sent and received. Live HTTP Headers can be found at https://addons.mozilla.org/en-US/firefox/addon/3829/ or by googling "Live HTTP Headers".

The role of HTTP Headers

Along with each HTTP request and response comes a bunch of headers carrying metadata. Varnish will look at these headers to determine if it is appropriate to cache the contents and how long Varnish can keep the content cached.

Please note that when Varnish considers these headers Varnish actually considers itself part of the actual webserver. The rationale being that both are under your control.

The term surrogate origin cache is not really well defined by the IETF or RFC 2616 so the various ways Varnish works might differ from your expectations.

Let's take a look at the important headers you should be aware of:

Cookies

Varnish will, in the default configuration, not cache an object coming from the backend with a 'Set-Cookie' header present. Also, if the client sends a Cookie header, Varnish will bypass the cache and go directly to the backend.

This can be overly conservative. A lot of sites use Google Analytics (GA) to analyze their traffic. GA sets a cookie to track you. This cookie is used by the client side javascript and is therefore of no interest to the server.

Cookies from the client

For a lot of web applications it makes sense to completely disregard the cookies unless you are accessing a special part of the web site. This VCL snippet in vcl_recv will disregard cookies unless you are accessing /admin/:

if (!(req.url ~ "^/admin/")) {
    unset req.http.Cookie;
}

Quite simple. If, however, you need to do something more complicated, like removing one out of several cookies, things get difficult. Unfortunately Varnish doesn't have good tools for manipulating the Cookies. We have to use regular expressions to do the work. If you are familiar with regular expressions you'll understand whats going on. If you aren't we recommend that you either pick up a book on the subject, read through the pcrepattern man page, or read through one of many online guides.

Lets use the Varnish Software (VS) web as an example here. Very simplified the setup VS uses can be described as a Drupal-based backend with a Varnish cache in front. VS uses some cookies for Google Analytics tracking and similar tools. The cookies are all set and used by Javascript. Varnish and Drupal doesn't need to see those cookies and since Varnish will cease caching of pages when the client sends cookies Varnish will discard these unnecessary cookies in VCL.

In the following VCL we discard all cookies that start with an underscore:

# Remove has_js and Google Analytics __* cookies.
set req.http.Cookie = regsuball(req.http.Cookie, "(^|;\s*)(_[_a-z]+|has_js)=[^;]*", "");
# Remove a ";" prefix, if present.
set req.http.Cookie = regsub(req.http.Cookie, "^;\s*", "");

Lets look at an example where we remove everything except the cookies named "COOKIE1" and "COOKIE2" and you can marvel at the "beauty" of it:

sub vcl_recv {
    if (req.http.Cookie) {
        set req.http.Cookie = ";" + req.http.Cookie;
        set req.http.Cookie = regsuball(req.http.Cookie, "; +", ";");
        set req.http.Cookie = regsuball(req.http.Cookie, ";(COOKIE1|COOKIE2)=", "; \1=");
        set req.http.Cookie = regsuball(req.http.Cookie, ";[^ ][^;]*", "");
        set req.http.Cookie = regsuball(req.http.Cookie, "^[; ]+|[; ]+$", "");

        if (req.http.Cookie == "") {
            unset req.http.Cookie;
        }
    }
}

A somewhat simpler example that can accomplish almost the same functionality can be found below. Instead of filtering out "other" cookies it instead picks out "the one" cookie that is needed, copies it to another header and then copies it back to the request, deleting the original cookie header.

sub vcl_recv {
    # save the original cookie header so we can mangle it
    set req.http.X-Varnish-PHP_SID = req.http.Cookie;
    # using a capturing sub pattern, extract the continuous string of
    # alphanumerics that immediately follows "PHPSESSID="
    set req.http.X-Varnish-PHP_SID =
       regsuball(req.http.X-Varnish-PHP_SID, ";? ?PHPSESSID=([a-zA-Z0-9]+)( |;| ;).*","\1");
    set req.http.Cookie = req.X-Varnish-PHP_SID;
    unset req.X-Varnish-PHP_SID;
}

There are other scary examples of what can be done in VCL in the Varnish Cache Wiki.

Cookies coming from the backend

If your backend server sets a cookie using the 'Set-Cookie' header Varnish will not cache the page when using the default configuration. A hit-for-miss object (see actions) is created. So, if the backend server acts silly and sets unwanted cookies just unset the 'Set-Cookie' header and all should be fine.

Cache-Control

The 'Cache-Control' header instructs caches how to handle the content. Varnish cares about the max-age parameter and uses it to calculate the TTL for an object.

So make sure you issue a 'Cache-Control' header with a max-age header. You can have a look at what Varnish Software's Drupal server issues:

$ GET -Used http://www.varnish-software.com/|grep ^Cache-Control
Cache-Control: public, max-age=600

Age

Varnish adds an 'Age' header to indicate how long the object has been kept inside Varnish. You can grep out 'Age' from varnishlog with varnishlog -I RespHeader:^Age.

Pragma

An HTTP 1.0 server might send the header Pragma: nocache. Varnish ignores this header. You could easily add support for this header in VCL.

In vcl_backend_response:

if (beresp.http.Pragma ~ "nocache") {
    set beresp.uncacheable = true;
    set beresp.ttl = 120s; # how long not to cache this url.
}

Authorization

If Varnish sees an 'Authorization' header it will pass the request. If this is not what you want you can unset the header.

Overriding the time-to-live (TTL)

Sometimes your backend will misbehave. It might, depending on your setup, be easier to override the TTL in Varnish then to fix your somewhat cumbersome backend.

You need VCL to identify the objects you want and then you set the 'beresp.ttl' to whatever you want:

sub vcl_backend_response {
    if (bereq.url ~ "^/legacy_broken_cms/") {
        set beresp.ttl = 5d;
    }
}

This example will set the TTL to 5 days for the old legacy stuff on your site.

Forcing caching for certain requests and certain responses

Since you still might have this cumbersome backend that isn't very friendly to work with you might want to override more stuff in Varnish. We recommend that you rely as much as you can on the default caching rules. It is perfectly easy to force Varnish to lookup an object in the cache but it isn't really recommended.

Normalizing your namespace

Some sites are accessed via lots of hostnames. http://www.varnish-software.com/, http://varnish-software.com/ and http://varnishsoftware.com/ all point at the same site. Since Varnish doesn't know they are the same, Varnish will cache different versions of every page for every hostname. You can mitigate this in your web server configuration by setting up redirects or by using the following VCL:

if (req.http.host ~ "(?i)^(www.)?varnish-?software.com") {
    set req.http.host = "varnish-software.com";
}

HTTP Vary

HTTP Vary is not a trivial concept. It is by far the most misunderstood HTTP header.

A lot of the response headers tell the client something about the HTTP object being delivered. Clients can request different variants of a HTTP object, based on their preference. Their preferences might cover stuff like encoding or language. When a client prefers UK English this is indicated through Accept-Language: en-uk. Caches need to keep these different variants apart and this is done through the HTTP response header 'Vary'.

When a backend server issues a Vary: Accept-Language it tells Varnish that its needs to cache a separate version for every different Accept-Language that is coming from the clients.

If two clients say they accept the languages "en-us, en-uk" and "da, de" respectively, Varnish will cache and serve two different versions of the page if the backend indicated that Varnish needs to vary on the 'Accept-Language' header.

Please note that the headers that 'Vary' refer to need to match exactly for there to be a match. So Varnish will keep two copies of a page if one of them was created for "en-us, en-uk" and the other for "en-us,en-uk". Just the lack of a whitespace will force Varnish to cache another version.

To achieve a high hitrate whilst using Vary is there therefore crucial to normalize the headers the backends varies on. Remember, just a difference in casing can force different cache entries.

The following VCL code will normalize the 'Accept-Language' header to either "en", "de" or "fr", in this order of precedence:

if (req.http.Accept-Language) {
    if (req.http.Accept-Language ~ "en") {
        set req.http.Accept-Language = "en";
    } elsif (req.http.Accept-Language ~ "de") {
        set req.http.Accept-Language = "de";
    } elsif (req.http.Accept-Language ~ "fr") {
        set req.http.Accept-Language = "fr";
    } else {
        # unknown language. Remove the accept-language header and
        # use the backend default.
        unset req.http.Accept-Language
    }
}

Vary parse errors

Varnish will return a "503 internal server error" page when it fails to parse the 'Vary' header, or if any of the client headers listed in the Vary header exceeds the limit of 65k characters. An 'SLT_Error' log entry is added in these cases.

Pitfall - Vary: User-Agent

Some applications or application servers send Vary: User-Agent along with their content. This instructs Varnish to cache a separate copy for every variation of 'User-Agent' there is and there are plenty. Even a single patchlevel of the same browser will generate at least 10 different 'User-Agent' headers based just on what operating system they are running.

So if you really need to vary based on 'User-Agent' be sure to normalize the header or your hit rate will suffer badly. Use the above code as a template.

Cache misses

When Varnish does not find an object for a request in the cache, then by default it performs a fetch from the backend on the hypothesis that the response might be cached. This has two important consequences:

  • Concurrent backend requests for the same object are coalesced -- only one fetch is executed at a time, and the other pending fetches wait for the result (unless you have brought about one of the states described below in Uncacheable content). This is to prevent your backend from being hit by a "thundering herd" when the cached response has expired, or if it was never cached in the first place. If it turns out that the response to the first fetch is cached, then that cache object can be delivered immediately to other pending requests.
  • The backend request for the cache miss cannot be conditional if Varnish does not have an object in the cache to validate; that is, it cannot contain the headers If-Modified-Since or If-None-Match, which might cause the backend to return status "304 Not Modified" with no response body. Otherwise, there might not be a response to cache. If those headers were present in the client request, they are removed from the backend request.

By setting a grace time for cached objects (default 10 seconds), you allow Varnish to serve stale content while waiting for coalesced fetches, which are run asynchronously while the stale response is sent to the client. For details see Misbehaving servers.

Although the headers for a conditional request are removed from the backend fetch on a cache miss, Varnish may nevertheless respond to the client request with "304 Not Modified" if the resulting response allows it. At delivery time, if the client request had an If-None-Match header that matches the ETag header in the response, or if the time in an If-Modified-Since request header is equal to or later than the time in the Last-Modified response header, Varnish will send the 304 response to the client. This happens for both hits and misses.

Varnish can send conditional requests to the backend if it has an object in the cache against which the validation can be performed. You can ensure that an object is retained for this purpose by setting beresp.keep in vcl_backend_response:

sub vcl_backend_response {
  # Keep the response in cache for 4 hours if the response has
  # validating headers.
  if (beresp.http.ETag || beresp.http.Last-Modified) {
    set beresp.keep = 4h;
  }
}

A stale object is not removed from the cache for the duration of beresp.keep after its TTL and grace time have expired. This will increase the storage requirements for your cache, but if you have the space, it might be worth it to keep stale objects that can be validated for a fairly long time. If the backend can send a 304 response long after the TTL has expired, you save bandwith on the fetch and reduce pressure on the storage; if not, then it's no different from any other cache miss.

If, however, you would prefer that backend fetches are not conditional, just remove the If-* headers in vcl_backend_fetch:

sub vcl_backend_fetch {
  # To prevent conditional backend fetches.
  unset bereq.http.If-None-Match;
  unset bereq.http.If-Modified-Since;
}

That should only be necessary if the conditional fetches are problematic for the backend, for example if evaluating whether the response is unchanged is too costly for the backend app, or if the responses are just buggy. From the perspective of Varnish, 304 responses are clearly preferable; fetches with the empty response body save bandwidth, and storage does not have to be allocated in the cache, since the existing cache object is re-used.

To summarize, you can improve performance even in the case of cache misses by:

  • ensuring that cached objects have a grace time during which a stale object can be served to the client while fetches are performed in the background, and
  • setting a keep time for cached objects that can be validated with a 304 response after they have gone stale.

Uncacheable content

Some responses cannot be cached, for various reasons. The content may be personalized, depending on the content of the Cookie header, or it might just be the sort of thing that is generated anew on each request. The cache can't help with that, but nevertheless there are some decisions you can make that will help Varnish deal with uncacheable responses in a way that is best for your requirements.

The issues to consider are:

  • preventing request coalescing
  • whether (and how soon) the response for the same object may become cacheable again
  • whether you want to pass along If-Modified-Since and If-None-Match headers from the client request to the backend, to allow the backend to respond with status 304

Passing client requests

Depending on how your site works, you may be able to recognize a client request for a response that cannot be cached, for example if the URL matches certain patterns, or due to the contents of a request header. In that case, you can set the fetch to pass with return(pass) from vcl_recv:

sub vcl_recv {
  if (req.url ~ "^/this/is/personal/") {
    return(pass);
  }
}

For passes there is no request coalescing. Since pass indicates that the response will not be cacheable, there is no point in waiting for a response that might be cached, and all pending fetches for the object are concurrent. Otherwise, fetches waiting for an object that turns out to be uncacheable after all may be serialized -- pending fetches would wait for the first one, and when the result is not entered into the cache, the next fetch begins while all of the others wait, and so on.

When a request is passed, this can be recognized in the vcl_backend_* subroutines by the fact that bereq.uncacheable and beresp.uncachable are both true. The backend response will not be cached, even if it fulfills conditions that otherwise would allow it, for example if Cache-Control sets a positive TTL.

Pass is the default (that is, builtin.vcl calls return(pass) in vcl_recv) if the client request meets these conditions:

  • the request method is a standard HTTP/1.1 method, but not GET or HEAD
  • there is either a Cookie or an Authorization header, indicating that the response may be personalized

If you want to override the default, say if you are certain that the response may be cacheable despite the presence of a Cookie, make sure that a return gets called at the end of any path that may be taken through your own vcl_recv. But if you do that, no part of the built-in vcl_recv gets executed; so take a close look at vcl_recv in builtin.vcl, and duplicate any part of it that you require in your own vcl_recv.

As with cache hits and misses, Varnish decides to send a 304 response to the client after a pass if the client request headers and the response headers allow it. This might mean that Varnish will send a 304 response to the client even after the backend saw the same request headers (If-Modified-Since and/or If-None-Match), but decided not to respond with status 304, while nevertheless setting the response headers ETag and/or Last-Modified so that 304 would appear to be warranted. If you would prefer that Varnish doesn't do that, then remove the If-* client request headers in vcl_pass:

sub vcl_pass {
  # To prevent 304 client responses after a pass.
  unset req.http.If-None-Match;
  unset req.http.If-Modified-Since;
}

hit-for-miss

You may not be able to recognize all requests for uncacheable content in vcl_recv. You might want to allow backends to determine their own cacheability by setting the Cache-Control header, but that cannot be seen until Varnish receives the backend response, so vcl_recv can't know about it.

By default, if a request is not passed and the backend response turns out to be uncacheable, the cache object is set to "hit-for-miss", by setting beresp.uncacheable to true in vcl_backend_response. A minimal object is saved in the cache, so that the "hit-for-miss" state can be recognized on subsequent lookups. (The cache is used to remember that the object is uncacheable, for a limited time.) In that case, no request coalescing is performed, so that fetches can run concurrently. Otherwise, fetches for hit-for-miss are just like cache misses, meaning that:

  • the response may become cacheable on a later request, for example if it sets a positive TTL with Cache-Control, and
  • fetches cannot be conditional, so If-Modified-Since and If-None-Match headers are removed from the backend request.

When beresp.uncacheable is set to true, then beresp.ttl determines how long the hit-for-miss state may last at most. The hit-for-miss state ends after this period of time elapses, or if a cacheable response is returned by the backend before it elapses (the elapse of beresp.ttl just means that the minimal cache object expires, like any other cache object expiration). If a cacheable response is returned, then that object replaces the hit-for-miss object, and subsequent requests for it will be cache hits. If no cacheable response is returned before beresp.ttl elapses, then the next request for that object will be an ordinary miss, and hence will be subject to request coalescing.

When Varnish sees that it has hit a hit-for-miss object on a new request, it executes vcl_miss, so any custom VCL you have written for cache misses will apply in the hit-for-miss case as well.

builtin.vcl sets beresp.uncacheable to true, invoking the hit-for-miss state, under a number of conditions that indicate that the response cannot be cached, for example if the TTL was computed to be 0 or if there is a Set-Cookie header. beresp.ttl is set to two minutes by builtin.vcl in this case, so that is how long hit-for-miss lasts by default.

You can set beresp.uncacheable yourself if you need hit-for-miss on other conditions:

sub vcl_backend_response {
  if (beresp.http.X-This-Is == "personal") {
    set beresp.uncacheable = true;
  }
}

Note that once beresp.uncacheable has been set to true it cannot be set back to false; attempts to do so in VCL are ignored.

Although the backend fetches are never conditional for hit-for-miss, Varnish may decide (as in all other cases) to send a 304 response to the client if the client request headers and response headers ETag or Last-Modified allow it. If you want to prevent that, remove the If-* client request headers in vcl_miss:

sub vcl_miss {
  # To prevent 304 client responses on hit-for-miss.
  unset req.http.If-None-Match;
  unset req.http.If-Modified-Since;
}

hit-for-pass

A consequence of hit-for-miss is that backend fetches cannot be conditional, since hit-for-miss allows subsequent responses to be cacheable. This may be problematic for responses that are very large and not cacheable, but may be validated with a 304 response. For example, you may want clients to validate an object via the backend every time, only sending the response when it has been changed.

For a situation like this, you can set an object to "hit-for-pass" with return(pass(DURATION)) from vcl_backend_response, where the DURATION determines how long the hit-for-pass state lasts:

sub vcl_backend_response {
  # Set hit-for-pass for two minutes if TTL is 0 and response headers
  # allow for validation.
  if (beresp.ttl <= 0s && (beresp.http.ETag || beresp.http.Last-Modified)) {
    return(pass(120s));
  }
}

As with hit-for-miss, a minimal object is entered into the cache so that the hit-for-pass state is recognized on subsequent requests. The request is then processed as a pass, just as if vcl_recv had returned pass. This means that there is no request coalescing, and that If-Modified-Since and If-None-Match headers in the client request are passed along to the backend, so that the backend response may be 304.

Varnish executes vcl_pass when it hits a hit-for-pass object. So again, you can arrange for your own handling of both pass and hit-for-pass with the same code in VCL.

If you want to prevent Varnish from sending conditional requests to the backend, then remove the If-* headers from the backend request in vcl_backend_fetch, as shown above for cache misses. And if you want to prevent Varnish from deciding at delivery time to send a 304 response to the client based on the client request and response headers, then remove the headers from the client request in vcl_pass, as shown above for pass.

The hit-for-pass state ends when the "hit-for-pass TTL" given in the return statement elapses. As with passes, the response to a hit-for-pass fetch is never cached, even if it would otherwise fulfill conditions for cacheability. So unlike hit-for-miss, it is not possible to end the hit-for-pass state ahead of time with a cacheable response. After the "hit-for-pass TTL" elapses, the next request for that object is handled as an ordinary miss.

It is possible to end the hit-for-pass state of a cache object by setting req.hash_always_miss to true in vcl_recv for a request that will hit the object (you'll have to write VCL that brings that about). The request in which that happens is forced to be a cache miss, and the state of the object afterwards depends on the disposition of the backend response -- it may become a cache hit, hit-for-miss, or may be set to hit-for-pass again.

hit-for-miss is the default treatment of uncacheable content. No part of builtin.vcl invokes hit-for-pass, so if you need it, you have to add the necessary return statement to your own VCL.