Hi everyone,
Apologies if this is really not VPS-related, but I thought I'd post it
just in case it might be...
I maintain the clock on my VPS (running Debian Lenny) with ntpd however I
am having an occasional issue with it insofar that every couple of months
or so it is suddenly making a massive timeshift (-46 seconds in this
case):
Sep 6 12:29:29 targur ntpd[1412]: kernel time sync status change 4001
Sep 6 12:46:34 targur ntpd[1412]: kernel time sync status change 0001
Sep 6 13:55:40 targur ntpd[1412]: synchronized to 74.118.152.85, stratum 2
Sep 6 13:56:37 targur ntpd[1412]: no servers reachable
Sep 6 14:07:10 targur ntpd[1412]: synchronized to 80.85.129.103, stratum 2
Sep 6 14:07:52 targur ntpd[1412]: time reset -46.871472 s
Sep 6 14:07:57 targur ntpd[1412]: synchronized to 209.237.247.192, stratum 3
Sep 6 14:08:09 targur ntpd[1412]: synchronized to
2001:7b8:633:1:213:154:236:182, stratum 2
Sep 6 14:12:10 targur ntpd[1412]: synchronized to 80.85.129.103, stratum 2
Sep 6 14:18:40 targur ntpd[1412]: synchronized to
2001:7b8:633:1:213:154:236:182, stratum 2
Sep 6 14:22:51 targur ntpd[1412]: synchronized to 80.85.129.103, stratum 2
Sep 6 15:48:24 targur ntpd[1412]: kernel time sync status change 4001
Sep 6 16:05:27 targur ntpd[1412]: kernel time sync status change 0001
Sep 6 17:30:53 targur ntpd[1412]: kernel time sync status change 4001
Sep 6 17:47:56 targur ntpd[1412]: kernel time sync status change 0001
Sep 6 18:56:11 targur ntpd[1412]: synchronized to
2001:7b8:633:1:213:154:236:182, stratum 2
The servers listed in /etc/ntp.conf are:
server ntp0.sfo.bitfolk.com iburst
server ntp0.lon.bitfolk.com iburst
server ntp1.lon.bitfolk.com iburst
server 0.us.pool.ntp.org iburst
server 1.us.pool.ntp.org iburst
server 2.us.pool.ntp.org iburst
server 0.pool.ntp.org iburst
server 1.pool.ntp.org iburst
server 2.pool.ntp.org iburst
Whilst the clock seems to get back on track the problem is that my IMAP
server throws a wobbly and self-destructs:
Sep 6 14:07:52 targur dovecot: Time just moved backwards by 47 seconds.
This might cause a lot of problems, so I'll just kill myself now.
http://wiki.dovecot.org/TimeMovedBackwards
...which whilst only requiring a simple restart of Dovecot I still have to
spot that it has happened first.
I am aware that I can tweak ntpd to not attempt such large jumps however I
can't help be feel that doing this is treating the symptom rather than the
cause and so I am keen to find out what the real problem is (it also has
other drawbacks such a slowing down corrective action when actually
required).
Does anyone else have this problem? I could understand such large jumps
following a fresh boot, or after a manual clock change, but not when
running in steady state...
Regards,
Mathew
Hi Andy,
At the time, I was running a Bitfolk VPS with 480MB RAM, but with rarely
more than 4 people online at the same time. The problem seemed to be less
related to the number of players and more to the size of the map over time -
the more people spread out and built things, the more had to be loaded in
memory while the server was running, as, for instance, just two people
standing at opposite ends of the world mean that two huge chunks of the
world need to be loaded at the same time. On a new, clean map, problems
were barely noticeable; after a month of play from just 4-6 people, the
entire VPS was barely responsive while Minecraft was running, and the
Minecraft server swamped the entire available RAM and swap even with just
one person online in a well-built area. I don't know enough about how
Multiplay set up their servers to fully know the hardware differences
involved, but running a huge, well-developed map on a 1GB RAM Multiplay
server presents no problems, while running the same map now on my Bitfolk
VPS at half the RAM would almost certainly have proved impossible.
The numbers provided for maximum players per RAM size struck me as a touch
inadequate, knowing how much RAM the same map had required when running on
my VPS, which is why we chose to to go for 1GB instead of 600MB. While
they're not cheap, and it's certainly possible you could undercut them with
more information on the technical requirements (I suspect you're right about
IOPS), the differences in noticeable performance are quite staggering, which
suggests that the hardware requirements for a well-developed map are likely
a little on the ridiculous side.
I'd be willing to help out with some interesting diagnostic testing on a
server if it'd help establish these sorts of requirements more concretely.
I know I'd personally rather have a Minecraft server running on a Bitfolk
VPS than a Multiplay one - for all their slickness, I miss being able to
roll a new Bukkit build when it's released and test experimental plugins,
rather than waiting for someone to pull their finger out and install the
latest Recommended Build.
Ta,
Tom
On Tue, Sep 6, 2011 at 6:07 AM, Andy Smith <andy(a)bitfolk.com> wrote:
> Hi Tom,
>
> Thanks for your reply, I was hoping you would chip in given your
> experience here.
>
> On Mon, Sep 05, 2011 at 09:18:32PM +0100, Tom Crosby wrote:
> > Honestly, you're really going to struggle to run anything for more than a
> > couple of people, and once you've done a decent bit of exploring and
> > building it's going to become unplayable unless you've really opted for a
> > reasonably huge amount of RAM - and then you're looking at quite an
> > expensive VPS for the purposes of running one game. I ran a server for a
> > handful of people for quite some time on a Bitfolk VPS
>
> Can you say how much RAM your VPS has, and how many Minecraft
> players you could comfortably have online before performance became
> unacceptable?
>
> > it was painful enough to convince us to fork out a little bit for
> > a Multiplay server instead.
>
> Looking at multiplaygameservers.com they seem to suggest that you
> should be able to fit the following amounts of players into these
> amounts of memory dedicated to the Minecraft server:
>
> Memory (MiB) | Max # of players
> -------------+-----------------
> 200 | 4
> 600 | 17
> 1024 | 35
> 2048 | 53
>
> Now, it does say that this is the absolute minimum memory, and that
> large maps may require more. Also bear in mind that they appear to
> be hosting off of SSDs which will likely be providing more IOPS than
> a BitFolk VPS.
>
> Can you say how much memory your Multiplay server has, and how many
> players you're able to support?
>
> It strikes me that 600MiB is not a huge amount of memory to dedicate
> to Minecraft and that 17 players may be plenty for some people. The
> IOPS may be an issue however.
>
> > sort of VPS setup that's likely to be affordable - you're much better off
> > getting something that's far less flexible but far more geared towards
> the
> > purpose.
>
> I've no idea how comparable a 720MiB VPS would be to a 600MiB
> Multiplay server, but the VPS will set you back £137.88/year inc.
> VAT, whereas the Multiplay server would be £153.00/year inc. VAT.[1]
>
> Conversely, for £167.88/year inc VAT you can have a 960MiB VPS.
>
> I'd be interested in finding out what the IOPS requirements are for
> a Minecraft server. I suspect they are quite high otherwise Multiplay
> wouldn't be using SSDs.
>
> Multiplay's prices seem quite high given that what you appear to get
> is a VPS dedicated to Minecraft, backed by SSD. Possibly I'm
> missing something. If not then I'm confident that I could undercut
> them with servers I have taken out of service for being too slow for
> VPS hosting, but with disks replaced by an SSD or two.
>
> The problem is that I don't know what the shelf life of Minecraft
> is, and I haven't got time to be playing computer games to work out
> what I could host next, nor to be replicating Multiplay's quite
> slick interface. I'd do all the work and then most likely Minecraft
> would become old hat and no one would want to buy servers for it.
> i.e. I don't see a way for me to make a product out of this.
>
> There's spare memory capacity at the moment. If anyone would like
> to experiment with how usable a Minecraft server is on a BitFolk VPS
> at various levels of RAM then I'd be willing to let you do that
> for free provided you write up your findings.
>
> Such VPSes would be purely for testing Minecraft though and once I
> need to sell the resources I'd have to convert you to a paying
> customer or end the arrangement.
>
> Also at the moment I am exploring caching block devices with SSDs.
> Once I have a server in that configuration colocated then it would
> be worth repeating the experiments to see if/how that improves
> matters.
>
> Finally, I might be willing to provide a single BitFolk Minecraft
> server free of charge for use by BitFolk customers (only), if any of
> you would actually use it. I personally probably wouldn't have time
> and I'm aware that several of you are running your own Minecraft
> servers already, and realise you probably want to focus your
> Minecraft attentions there. :)
>
> Cheers,
> Andy
>
> [1] I gather from
> http://forums.multiplay.co.uk/feedback-bug-reports-etc/78738-vat
> that Multiplay are not providing customers with VAT invoices, so
> those outside the EU are paying VAT when they don't need to, and
> VAT registered customers may find it difficult to claim back the
> VAT. Not that I imagine there are many business users of a
> Minecraft server...
>
> --
> http://bitfolk.com/ -- No-nonsense VPS hosting
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.10 (GNU/Linux)
>
> iEYEAREDAAYFAk5lqoEACgkQIJm2TL8VSQt1twCdH/FrwqcIfcbJSiKZ8ChWVmXs
> 2cIAniLR2yFezMXz8uRQhF99fr3Kl2Yh
> =UbJX
> -----END PGP SIGNATURE-----
>
>
> _______________________________________________
> users mailing list
> users(a)lists.bitfolk.com
> https://lists.bitfolk.com/mailman/listinfo/users
>
>
I've just had an interesting problem turn up: I replied to an email
on a kernel.org mailing list. There were Chinese characters near the
top of the original mail. On its way out through my mail hub, the mail
was run through the Bitfolk spamassassin, which added its usual two
headers: one with the spam score in it, and one with a horribly
mangled extract of the original mail. This mail got rejected by
vger.kernel.org on the grounds that it doesn't accept UTF-8 in mail
headers.
Is there any way I can get the X-frost.carfax.org.uk-Spam-Report:
header either suppressed completely, or (in preference) without the
content of the original message in it?
I'm guessing the answer is "not without running your own
spamassassin", since this is a shared service, but I thought I'd ask
anyway, just in case...
Hugo.
--
=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
--- This year, I'm giving up Lent. ---
Hello peeps. New here (looks around nervously).
I'm migrating from Gandi (gandi.fr, somewhere near Paris) and
was a bit surprised that my "rsync -azP" seemed to be shifting only
about one megabyte a second. Any ideas? AFAIK both ends are
sitting on fairly fat pipes so I would have hoped for a bit
better.
