Closed Bug 105746 Opened 23 years ago Closed 23 years ago

spurious, intermittent "not found" problems

Categories

(Core :: Networking, defect)

Product:
Core
Component:
Networking
Platform:
x86
All
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED INVALID
Milestone:
mozilla0.9.8

People

(Reporter: akkzilla, Assigned: gordon)

References

Details

Attachments

(2 files)

Screendump of message in mozilla 0.9.5/WinNT
54.56 KB, image/jpeg
Details
Screendump of error
4.89 KB, image/png
Details
I've been seeing intermittent "not found" on loading pages for about 2 weeks, maybe slightly longer. Here's an IRC conversation with details from other people who are also seeing this. <akk> I keep getting "not found" errors from mozilla, where trying a second time works. <Matti_> akk: wfm <HJ> akk, page loads? <akk> I see this a lot when going back in history -- go back to a page that was fine, get a "not found", <akk> hit reload, it's fine. <HJ> yep, same for me <jkeiser> akk: I'm seeing it too. It seems to be cache-related. <HJ> jkeiser, but I don't use cache at all! <ssieb_home> akk: yes, if you go forward and then back again, it usually works <akk> Anyone know if this is reported? Should I just remove my cache? <HJ> won't help <jkeiser> HJ: well, then maybe it's not :) Anyway, the intermittency of it seems to indicate cache to me <jkeiser> I've tried that too <ssieb_home> is it always on submitted pages? * HJ checks his cache settings <akk> No, not always submitted pages (usually not). <jkeiser> ssieb_home: nope <akk> I see this on about one page out of ten. <HJ> akk, I was trying to find out what exactly is going on, but no luck so far (:
Here's an example, where it happened not for a main page but for an ad included on that page -- I was going back (via history back button) to a page on ubid.com, maybe 5-10 minutes after visiting it, after viewing some other unrelated pages, and I got a popup: ubid.adbureau.net could not be found. Please check the name and try again. I've also seen it on bugzilla bug pages, and some people have seen it on mozilla.org pages (not sure if I have).
Darin suggested the possibility that this might be a DNS problem--specifically a problem that occurs when Mozilla tries to re-resolve a previously cached DNS entry. All of us seeing this are on Linux (different flavors). All of us but one have cache set to Automatic (and the other is not sure whether it was set there when he had problems). A way to quickly test and reproduce this would be to turn down the length of time DNS entries are cached for.
On Darin's suggestion, I tried setting the pref: user_pref("network.dnsCacheExpiration", 65,000); I just saw the symptoms again, curiously again on ubid (but honest, I've seen it on lots of other servers too!) I typed ubid.com into the urlbar, hit return, and got a popup saying: www.ubid.com could not be found. Please check the name and try again. It had been maybe 30-45 minute since I last went to anything on ubid.com (a lot shorter than 65000 seconds).
I set the pref to 1 and couldn't reproduce this. Incidentally, I haven't reproduced it since I set it to 30,000 either.
cc'ing gordon since the DNS cache seems to be suspect here.
After not seeing this all day (with network.dnsCacheExpiration set to 65,000) I'm suddenly seeing it repeatedly. I'm trying to load the sequential pages under http://www.timocharis.com/honeymoon/monday/ (which I had looked at maybe 3-5 hours ago but not since), and each time I click next, I get "page cannot be found" popup, I dismiss the popup and hit reload, I get the next page's text without the inlined image (and with no error indication saying there should be an image there, but sometimes it shows me the alt text; it doesn't wait a long time, either, but loads the page quickly with no image), hit reload again, and finally I get the image (from the network, not cache -- it's obvious from the speed difference).
Restarting cured the "not found" problem, though curiously, the images I had just loaded (by hitting reload twice) are still not in the cache -- when I go back to the site after a restart, they're each downloading again. Perhaps we're also not writing to cache when we get into this mode.
It doesn't sound like a DNS problem to me. It says "page" cannot be found, not the server. We need to take a look at the memory and disk cache to see if these documents are getting stored. It would also be interesting to see the HTTP headers for the pages this occurs on.
hrm.. i suppose it might help to know where in the code this message is being generated. akkana: can you give the exact text of the popup dialog? i don't see "page cannot be found" anywhere in the mozilla source. % cd /path/to/mozilla % find . -type f -exec grep -iH 'page cannot be found' \{} \; turned up nothing.
Akkana asked me to change the OS field, I've set it to All, because it's obviously not limited to Linux only as you can see on the attached screenshot! BTW, a simple reload 'solved' this problem.
OS: Linux → All
And that particular error (NS_ERROR_UNKNOWN_HOST) is only generated in nsDNSService.cpp lines 836, 1783 and 1855. Line 836: DoSyncLookup generates it when PR_GetIPNodeByName fails for any reason. Perhaps it's failing for the wrong reasons? Line 1783: appears to be Mac-only. Line 1855: I have no idea what this is. A text search for that WSA* identifier turns up no other hits in all of Mozilla.
to gordon with love ;-)
Assignee: neeti → gordon
Target Milestone: --- → mozilla0.9.8
Targeting for necko stability milestone 0.9.8.
Holy garbonzos, I'm seeing this 3 times as often before after the last week or so, both on my home computer and at work. The worst was just today; just about every time I hit the Back button after looking around the Swing tutorial (http://java.sun.com/docs/books/tutorial/uiswing/overview/index.html) I get this. I do tend to leave this up for a while.
Attached image Screendump of error —
Now I'm getting "File not found" issues. Related?
I'm seeing The file "/" could not be found. Please check the location and try again. Also, on: http://lxr.mozilla.org/mozilla/source/docshell/base/appstrings.properties#21 lines 21 and 28 define the same variable twice. Isn't that a bit weird?
Jabbadoo, that's a major typo!
Doug and Darin collide!
*** Bug 114030 has been marked as a duplicate of this bug. ***
I've been seeing this for a similar period of time on Linux. The errors I see are definitely of the form "/ not found" or "/foo/bar/baz.html not found". I think the attachment showing a host not found is a separate, DNS issue. As the initial reporter said, hitting reload essentially always succeeds. I see this happen several times a day, on known good web sites.
I'm seeing more of this behavior in 2001120806's build. I saw a 'File / not found' dialog several times in a row trying to visit a particular site. My net link was fairly saturated with other traffic at the time, including a 50 meg ftp download in another Mozilla window.
It's especially serious when you have frames :) (1) load page, get file not found error (2) reload. Frameset comes up, get 3 file not found errors (3) reload. All is well. I have also noticed this problem on images (I believe it's this problem). They show up as though they do not exist and then reloading the page makes them show up again. Specifically, the site I have seen this on recently is http://java.sun.com/j2se/1.4/docs/api/index.html. Same behavior as before, it only happens if I leave a page for a while and come back. Works on frames--i.e. if two of the frames have been recently viewed but one has not been viewed for a while, I only get one not found error.
it sounds like this might be a cache interaction of some sort... perhaps a problem reading from the cache, or perhaps a problem validating cache entries. can anyone tell from looking at the status bar what state we might be in when the not-found alert pops up?
This does look like a cache problem. Today I was surfing at www.mozilla.org/party/ .. 2000 .. and then clicking on the various pages and then started getting the "file not found" error. Trying to go mozilla.org also gave the same error continously this time. No matter how many times I tried. I went into preferences and cleared my Disk Cache, and the problem went away. I think this is a good sign that it has something to do with cache.
Yes, I saw it continuously as well. Previously reload fixed the problem. Something before the 13th and after the 8th (I think) broke it more. Reopening the browser fixes the problem.
i know this bug is marked OS:ALL, but could those that are seeing this problem please restate what OS's you're seeing this with? thx!
Linux for me (Mandrake 8.1).
Linux here (Slackware 8)
-> RedHat Linux 7.2
My reports were from Linux builds. I saw this yesterday for all URLs in my 'back' list, but not for newly visited pages, so it looks like a cache issue to me as wel.
*** Bug 116895 has been marked as a duplicate of this bug. ***
I am seeing "file not found" errors almost hourly. I have not seen the host name variant that others have seen. This is fairly new - never saw it with 0.9.6, but I started seeing it with one of the nightlies in December, before 0.9.7 was released. I am seeing it with build 2001122108 on my desk top (RH 7.2) and with an earlier build (pre 0.9.7) on my laptop (RH 7.1). Could it be that multiple browser windows is contributing? I can't remember seeing it happen when I only had one browser window open, but I can't be sure. Clearing the disk cache always fixes it. I can't yet come up with a series of steps to make it happen, but it is not site specific.
I am pretty convinced that the "file not found" errors are related to multiple open browser windows. For the last few days I have tried to go with just one browser window open at a time. As long as I had only one open I never saw this condition. But when I went to a site that liked to open links in a new window, it did not take long before the "file not found" appeared when I clicked the back button and I had to clear the disk cache to keep going.
Bob: could you please give the exact text of the dialog error message? thx!
It is the same as shown in the attachment in comment #18, just substitute the path of the moment. The path in the message is the file portion URL of the page you are trying to go back to (typically) - that is, the URL less the "http://host.domain" part. Sorry I didn't make that clear.
so the "file not found" error box could be a cache problem. the error message is the path of the URL as noted, but we can only get this error if we're reading from the disk cache. so, it seems that it must be the case that a disk cache entry is somehow not present when we go to read it. i wonder if reducing the disk cache size to something small wouldn't cause this bug to show up more frequently. bob: what is your disk cache size set to?
My cache size is 50,000 KB, but it never gets anywhere near that. Right now it is at 6,776 KB (using du -sk on my Cache directory under ~/.mozilla). It never gets much bigger. But the quickest way to get it to happen is to open different browser windows and browse some sites (ones that aren't in yor cache at first) and then go back to a page that is in the cache. I can't make it happen on demand, but that is what will bring it on fastest. The early comments to this bug talked about "host.domain not found" popups. Are we talking about the same problem or are there 2 bugs here?
Hi, I have an easy way to reproduce that bug: open a mozilla session and go to www.microsoft.com (just for example! don't think anything wrong of me ;) open a second session and go to www.linux.com in the first session click on a link in the second do the same go back to the first session and check out the history.back-button - there comes the error! it didn't come when I wanted to click on another link on the first sessions page - well, not the first time I tried but the second *weird*!
frederick: what build are you using and on what platform? i couldn't get the dialog to come up when i followed those steps.
I see this bug still on Linux, Mozilla build 2002010606. I don't know if there are any reliable steps to reproduce, but it happens with notable frequency. Whatever is getting spoiled (cache?), it makes it impossible to go 'back' to anything in my back list. In my other Mozilla window, I have three things in my 'back' list, and they all became unaccessible at once, with messages like 'The file / cannot be found. Please check the location and try again.' or 'The file /cgi-bin/log_display.pl?log_name=access&lines=500&grep=gash2 cannot be found. Please check the location and try again.' each of those messages being verbatim transcriptions of the dialog. If I use 'ctrl-N' to open a new window, I am able to visit each of the three locations in first window's 'back' list without incident, either by using my bookmarks list to visit the pages, or by the 'home' page coming up. In whichever case, once I visit a URL in the new browser window, I can then return to my first window and successfully use the 'back' list to return to those URLs. If I don't visit a given URL in the new window, I can't visit the URL in my first window, through the 'back' and 'forward' button lists at all.. the same message above appears.
I use Mozilla build 2001122108 with SuSE Linux 7.3 on an AMD Athlon
Hm, on Linux 2002010821, popping up a new window does not necessarily let me visit the same URL's that are giving the 'not found' message in the first window. Perhaps some code has changed since Linux 2002010806, or perhaps it's related to the amount of interim cache activity?
Hmmmm. Frederick's technique in comment 41 is not triggering the problem for me - Red Hat 7.2, Mozilla 2001122108. I do see it happening, just not that easily. When I first started seeing the problem, I tried entering the URL into the location bar, but that just resulted in the same message. I am pretty sure I tried doing that in a new window too, but not certain. I will try it again. So far the only way out for me has been to clear the disk cache, or restart the browser.
I upgraded to build 2002010908 and I can still reproduce the bug the way I described!!!
any chance these problems are related to people running multiple instances of mozilla with the same profile? i'm just curious, because i don't understand how the disk cache is getting corrupted.
Yes, I do run multiple instances of Mozilla with the same profile. I'm running Gnome under linux, and when I right click on a URL in the terminal window, it spawns a new instance of Mozilla rather than inducing the existing Mozilla to spawn a new window. It isn't clear that this multiple instancing is directly related to the problem, as I do see the 'not found' messages start to appear often times when only one instance is running.
ok, interesting... running multiple instances of mozilla w/ the same profile is definitely very bad and could cause this problem. in particular, i am curious which of the two flavors of "not found" messages you are seeing when only one instance of mozilla is running? do you see the "file not found" or the "host not found" error message?
I also run two instances of mozilla on a regular basis (a release build for my daily browsing, and a debug build for work). For years this wasn't a problem; any idea what might have changed to make this problem suddenly appear? (NS4 solved the problem by making the second and subsequent instances read-only with respect to cache and prefs, which would be an easy and effective solution for moz; I'm not sure why we're not doing it.)
I always see the file not found variant. I don't believe I've ever had a persistent 'host not found' message from cache.
akkana: there's a bug filed to do something similar to protect the profile directory. it has never been ok to run two instances of mozilla using the same profile.
> Perhaps some code has changed since Linux 2002010806, or perhaps > it's related to the amount of interim cache activity? Hmm, bug 112564 was fixed on that date and that had a lot to do with cache, maybe that cleared it up (or made it harder to trigger it)?
Well, the 2002010806 Linux build actually had better behavior than the 2002010821 Linux build, in as much as the 0806 build allowed me to visit a given URL with a secondary window, which had the effect of clearing the problem in the primary window. In the 2002010821 build, visiting URL's in "secondary" windows does not fix up the cache corruption problem in the "primary".
well, when there's really a cache problem when 2 sessions are started, just make a cache-file for every session so there's no conflict! it's better than giving the second session no write-right to the cache...
Should I file a new bug for comment #19 in this bug? It still shows that mistake.
alex: yes please... assign it to embedding:docshell
Filed bug #119285 for comment #19, as directed.
Figures. It was already filed as bug 113924.
I just got this again; I too am seeing file ... cannot be found, not host not found. It happened in my release build after (but not during) running a debug build concurrently. The status bar on the main window didn't show anything useful: it still showed "Done (%f sec)" from the last url successfully loaded.
I'm marking this bug INVALID. It all appears to be interesting wreckage from using the same profile from multiple instances of the browser simulatneously (even if it's only starting up a browser and quiting). There is another bug (bug 76431) which details the need to protect profiles, so we don't need this one.
Status: NEW → RESOLVED
Closed: 23 years ago
Resolution: --- → INVALID
I guess no one is interested in understanding what changed to make this regress sometime in october? Before October, I'd been running multiple browser instances for over a year without ever seeing this problem.
akkana: it doesn't matter... you were just lucky... running multiple instances of mozilla w/ the same profile has never been a supported feature. in fact, the results of doing so have always been undefined. i don't think it's worth anyones time to try to figure out why it once worked w/ higher probability of success ;-)
Is this a depends or duplicate of bug 76431? There are lots of bad things with running multiple instances, which is really easy to do, but I don't see that bug being fixed quickly.
It could be considered a duplicate I suppose.
*** Bug 118760 has been marked as a duplicate of this bug. ***
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: