Skip to main content

dns resolver behavior

i had an occasion to have to look up windows client behavior when it came to dns. specifically, i wanted to know how the client behaves when the primary name server is offline. before i had to fire up packet trace and check for myself, i stumbled on a couple of useful articles that spell it out.

UPDATE: had a conversation with a talented linux dns guy and discovered a few more useful things to note.

dns client resolver behavior
dns client resolution timeouts
dns forwarders and conditional forwarders resolution timeouts

in summary, it works as follows:

  • dns query sent to preferred
  • if no response within 1 second, dns query sent to alternate
  • if no response within 1 second, dns query sent to preferred again
  • if no response within 2 seconds, dns query sent to preferred and alternate
  • if no response within 4 seconds, dns query sent to preferred and alternate again
  • if no response within 7 seconds, process times out

 

something to note for linux systems, these appear to be default values:

  • timeout:n
    sets the amount of time the resolver will wait for a response from a remote name server before retrying the query via a different name server.  Measured in seconds, the default is RES_TIMEOUT (currently 5, see <resolv.h>).  The value for this option is silently capped to 30.
  • attempts:n
    sets the number of times the resolver will send a query to its name servers before giving up and returning an error to the calling application.  The default is RES_DFLRETRY (currently 2, see <resolv.h>).  The value for this option is silently capped to 5.
  • rotate
    sets RES_ROTATE in _res.options, which causes round-robin selection of name servers from among those listed.  This has the effect of spreading the query load among all listed servers, rather than having all clients try the first listed server first every time.
  • search
    Resolver queries having fewer than ndots dots default is 1) in them will be attempted using each component of the search path in turn until a match is found.
  • ndots:n
    sets a threshold for the number of dots which must appear in a name given to res_query(3) (see resolver(3)) before an initial absolute query will be made.  The default for n is 1, meaning that if there are any dots in a name, the name will be tried first as an absolute name before any search list elements are appended to it.  The value for this option is silently capped to 15.

in summary, the timeout value indicates how long the client will wait until it tries the next server in the search list. the number of queries attempted per server is dependent on how ndots is configured. once the server list is exhausted, the attempts value indicates if the client should try the list again.

additional settings can be found here: http://man7.org/linux/man-pages/man5/resolver.5.html

Labels:

Comments

Post a Comment

Popular posts from this blog

using preloadpkgonsite.exe to stage compressed copies to child site distribution points

UPDATE: john marcum sent me a kind email to let me know about a problem he ran into with preloadpkgonsite.exe in the new SCCM Toolkit V2 where under certain conditions, packages will not uncompress.  if you are using the v2 toolkit, PLEASE read this blog post before proceeding.   here’s a scenario that came up on the mssms@lists.myitforum.com mailing list. when confronted with a situation of large packages and wan links, it’s generally best to get the data to the other location without going over the wire. in this case, 75gb. :/ the “how” you get the files there is really not the most important thing to worry about. once they’re there and moved to the appropriate location, preloadpkgonsite.exe is required to install the compressed source files. once done, a status message goes back to the parent server which should stop the upstream server from copying the package source files over the wan to the child site. anyway, if it’s a relatively small amount of packages, you can
11 comments
more »

How to Identify Applications Using Your Domain Controller

Problem Everyone has been through it. We've all had to retire or replace a domain controller at some point in our checkered collective experiences. While AD provides very intelligent high availability, some applications are just plain dumb. They do not observe site awareness or participate in locating a domain controller. All they want is the name or IP of one domain controller which gets hardcoded in a configuration file somewhere, deeply embedded in some file folder or setting that you are never going to find. How do you look at a DC and decide which applications might be doing it? Packet trace? Logs? Shut it down and wait for screaming? It seems very tedious and nearly impossible. Potential Solution Obviously I wouldn't even bother posting this if I hadn't run across something interesting. :) I ran across something in draftcalled Domain Controller Isolation. Since it's in draft, I don't know that it's published yet. HOWEVER, the concept is based off
3 comments
more »

sccm: content hash fails to match

back in 2008, I wrote up a little thing about how distribution manager fails to send a package to a distribution point . even though a lot of what I wrote that for was the failure of packages to get delivered to child sites, the result was pretty much the same. when the client tries to run the advertisement with an old package, the result was a failure because of content mismatch. I went through an ordeal recently capturing these exact kinds of failures and corrected quite a number of problems with these packages. the resulting blog post is my effort to capture how these problems were resolved. if nothing else, it's a basic checklist of things you can use.   DETECTION status messages take a look at your status messages. this has to be the easiest way to determine where these problems exist. unfortunately, it requires that a client is already experiencing problems. there are client logs you can examine as well such as cas, but I wasn't even sure I was going to have enough m
3 comments
more »