From: William S. <wst...@po...> - 2004-11-23 07:16:59
|
Good morning, all, (For the UML-devel mailing list: The "spamgate" virtual machine creates multiple text files and stores them on the host via a hostfs mount. The slartibartfast and jeeves VM's publish that data read through a hostfs mount out via http and rsync, respectively.) The three vm's were publishing just fine in the 2.4.23 kernel I have at http://www.stearns.org/uml/ . When I rebooted the three vm's into the 2.4.27 kernel I have there, slart and jeeves started showing stale data. Moving spamgate back to 2.4.23 didn't help; slart and jeeves still presented old files to the outside world. As soon as I moved slart and jeeves back to 2.4.23, the problems went away. Kernels and compile options used are both at http://www.stearns.org/uml/ . On Mon, 22 Nov 2004, Jeff Chan wrote: > On Monday, November 22, 2004, 10:25:15 PM, Alex Broens wrote: > > Jeff Chan wrote: > > > BTW, the big file transfer was after I manually deleted > >> sa-blacklist.current.domains on my end, but comparing a > >> version of sa-blacklist.current.domains saved last night > >> to the new version rsynced over, it's unchanged. > >> > >> (That's bad, unless the file is truly unchanged from about 10 > >> hours ago, which seem possible, but unlikely.) > > > Jeff, > > > added mirrorforhappiness. com a while ago. > > > Its in /processing/master-black now > > > it should show up in sa-blacklist.current.domains soon. > > > If I seee it in there, Id say you should as well, right? > > Yes, but the copy of sa-blacklist.current.domains I'm rsyncing > has not updated in almost 24 hours. You can tell by checking > the zone file serial numbers which are all in UNIX epoch seconds: > > http://www.surbl.org/nameservers-output.html > > The serial for ws.surbl.org 1101111492 is almost a day > old. That's possible if the data itself don't update > often, but I'm pretty sure it should be more often > than once a day! > > My script does not produce a new zone file unless > the data has changed: > > http://spamcheck.freeapp.net/handle-sa-blacklist > > Hmmm, the version I grabbed using http: > > lwp-request 'http://www.stearns.org/sa-blacklist/sa-blacklist.current.domains' > bar > > does not match the version I'm getting via rsync. > > But the http version matches the version I just rsynced into a > *new* file name: > > rsync -vv -L zaphod.stearns.org::wstearns/sa-blacklist/sa-blacklist.current.domains baz > > Bill can you check your server logs to see if there's anything > unusual. > > Is rsync somehow confused about timestamps? > > It definitely seems to be an rsync issue. Is there a place rsync > stores temporary files on your server or mine that might be > stale? Yes. :-) Cheers, - Bill --------------------------------------------------------------------------- The web page you seek cannot be found here: countless others await (Courtesy of John Sage <js...@fi...>) -------------------------------------------------------------------------- William Stearns (wst...@po...). Mason, Buildkernel, freedups, p0f, rsync-backup, ssh-keyinstall, dns-check, more at: http://www.stearns.org -------------------------------------------------------------------------- |
From: Blaisorblade <bla...@ya...> - 2004-11-23 17:31:35
|
On Tuesday 23 November 2004 08:17, William Stearns wrote: > Good morning, all, > (For the UML-devel mailing list: The "spamgate" virtual machine > creates multiple text files and stores them on the host via a hostfs > mount. The slartibartfast and jeeves VM's publish that data read through > a hostfs mount out via http and rsync, respectively.) > The three vm's were publishing just fine in the 2.4.23 kernel I > have at http://www.stearns.org/uml/ . When I rebooted the three vm's into > the 2.4.27 kernel I have there, slart and jeeves started showing stale > data. Moving spamgate back to 2.4.23 didn't help; slart and jeeves still > presented old files to the outside world. As soon as I moved slart and > jeeves back to 2.4.23, the problems went away. > Kernels and compile options used are both at > http://www.stearns.org/uml/ . =46rom the common experience, having hostfs working is a luck, with kernels= =20 later than 2.4.24-1um. 2.4.27-1bs (on my page) excludes all hostfs changes, while merging everythi= ng=20 else up to 2.4.26-3um, plus some other changes. Also, there are some unconfirmed results that NFS may be _faster_ than host= fs=20 =2D so you might want to experiment this. =2D-=20 Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 |