Closed Bug 117613 Opened 24 years ago Closed 22 years ago

DNS: Cached state of failed prior lookup supersedes available nameserver

Categories

(Core :: Networking, defect, P3)

defect

Tracking

()

VERIFIED INVALID
mozilla1.3alpha

People

(Reporter: Mozilla.20.TEN, Assigned: gordon)

Details

From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i586; en-US; rv:0.9.5) Gecko/20011012 BuildID: 2001101202 When the name server is not (yet) available (e.g. in a dial-up environment), Mozilla will report that the domain entered could not be resolved, and ask the user to retry at a later point in time. However, it will KEEP doing so (even after reducing the network.dnsCacheExpiration" down from its default of 5000) until the browser is closed and restarted (see Additional Information below for workaround), even when the name server has become available, i.e. even though the DNS lookup could then be carried out successfully. Reproducible: Always Steps to Reproduce: 1.Be sure not to have access to the name server (e.g. on ISDN dial-up without DoD, this would be the case after isdnctrl hangup ipppX) 2.Close and restart Mozilla 3.Try to access an URL that is known to exist and requiring to be resolved through DNS 4.Expected error message shows up 5.Now make sure system does get access to the name server (e.g. by isdnctrl dial ipppX in above setup, or by setting the appropriate route to the internet gateway) 6.Retry to access the URL chosen above 7.Wait for the timeout specified in your prefs.js to elapse 8.Retry to access the URL chosen above 9.Retry any other URL that is known to exist and requiring to be resolved through DNS Actual Results: The error message keeps reappearing even though the DNS is available on points 6, 8 and even 9 above. Expected Results: In the above cases of 6, 8 and, definitely, 9, the IP address not being known yet, it should have been resolved by retrying to query the DNS which had become available, rather than relying on a cache which does not hold a result yet Workaround: Other than closing and restarting the browser, when the connection to the Internet (and therefore the name server) has become available, clicking File/Work Offline, then File/Work Online will remedy the situation, i.e. the domain will be resolved properly after performing these steps.
*** This bug has been marked as a duplicate of 64857 ***
Status: UNCONFIRMED → RESOLVED
Closed: 24 years ago
Resolution: --- → DUPLICATE
Based on other comments, this sounds like a problem where DNS stops working if a DNS query fails to connect to the server.
Status: RESOLVED → UNCONFIRMED
Resolution: DUPLICATE → ---
Just want to add that I am able to reproduce that on Win95 using Mozilla 0.9.9. (directly entering the IP number instead works). The "Work offline/online" workaround does not work around anymore (as Al already stated for Linux in Comment #66 of Bug 64857). One difference I see is that with no connection during startup and first DNS lookup the only way to make Mozilla work again is restarting it, while everything works well after the dial-up connection has only been interrupted for some time and a DNS lookup has been tried during this interruption (of course without success). Even if the internet provider is a different one, DNS lookup works again after reconnecting. This means that I do NOT see "DNS stop working if a DNS query fails to connect to the server" (although I have to admit I also had this impression before I tested it minutes ago).
Enhancements to DNS service are coming, to make it smarter. Still, I think we need a button for the user, hence bug 140232.
Is this not just bug 117628?
Probably. I am going to re-read everything after I verify that fix, b/c I suspect there is more than one Linux-only problem.
just saw this yesterday (with 2002073022), added a host which DNS was broken to my /etc/hosts and it didn't help - other browsers had no problem, because they wheren't running before - so it can't be that FIXED bug ;)
Status: UNCONFIRMED → NEW
Ever confirmed: true
That might suggest that the DNS cache is not reset on failure, but we also have new DNS behavior, perhaps neeti could find out who would be the best person to explain those changes.
-->dns
Assignee: neeti → gordon
Summary: Cached state of failed prior DNS lookup supersedes available nameserver → DNS: Cached state of failed prior lookup supersedes available nameserver
I have been getting this bug for sometime. I hate this because I use notebook computer with both wireless and wired networkings. Everytime I switch networking, I need to relaunch Mozilla to access sites previously accessed via different networking. MacOS9.x 2002082008 trunk.
*** Bug 166146 has been marked as a duplicate of this bug. ***
Dupe was Mac/MacOS X -> All/All
OS: Linux → All
Hardware: PC → All
Priority: -- → P3
Target Milestone: --- → mozilla1.3alpha
I'm beginning to doubt that mozilla was ever made for use with DNSes. And no, going offline, then online, changing the DNSes, waiting 15 minutes or anything else doesn't work. Just tell me where this cache is so I can delete it.
It looks like the offline/online toggle via the File->Work Offline menu is no longer clearing the DNS cache and I consider that a regression. It shouldn't be too hard to fix. There are other places DNS results can get cached as well (glibc, socket transport service, PAC), and fixing the DNS service won't help or hinder those. If they still cause problems after this bug is fixed, we'll need to open new bugs.
Status: NEW → ASSIGNED
gordon: if you are going to fix that UI element, could you do it by reopening bug 192798, or filing a new (regression bug)? This bug really should be marked as a dupe of a "network.dnsCacheExpiration" or "I don't like your caching behavior" bug.
Okay Ben, I've reopened bug 192798, and added David Oftedal to the cc list (so he can see when we fix it). You can dup this one to whatever you want.
RESOLVED/dupe: This bug actually mixed up several problems, so I reopened the dupe and will clean that up separately. The original report was a linux user suffering from cached DNS servers (hence the appearnce of failed requests being cached). That is caused in part by our lack of good DNS error handling, bug 164715. The user's expectation that DNS cache should expire is bug 164715, but the actual situation never made it that far, so that is irrelevant. Since we don't have a negative cache, the summary is largely invalid (bug 164715 requests this as a feature). So, based on the summary, invalid.
Status: ASSIGNED → RESOLVED
Closed: 24 years ago22 years ago
Resolution: --- → INVALID
verified invalid
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.