Removing/Cemsoring bad Fedora mirrors?

[Dan Thurman]

I have in the past, complained of hanging yum
updates and over time, I finally realized that
there are such things as bad mirrors, and for some
reason or another, causes yum to hang indefinately.

One would think that yum can detect "hangs" and
move on to the next mirror to pick up the slack,
but this is not the case, or so I am led to believe...

Yes, I can add the bad mirror(s) to the fastestmirror.conf
in Yum's pluginconf.d directory, but then again,
why isn't it that these bad mirrors are being removed
for misbehaving?

For those who do manage mirror lists, the specific mirror
in question is: "web-ster.com"

Kind regards,
Dan

[Tom Horsley]

On Sat, 05 Jun 2010 10:27:15 -0700
Daniel B. Thurman wrote:

>  but that does not explain how the yum hang
>  remained - holding the connection "hostage" - that is the
>  part I cannot fathom.... 
I see these sorts of hangs all the time with lots of
testing going on on a big collection of both real and
virtual machines using NFS and ssh and wot-not. I don't
know exactly what the heck network timeouts timeout
on, but I often get things hung for hours with network
problems. Somehow networks go down but act like they are up :-).

I finally put independent fallback timeouts in my
test drivers to kill things off after they haven't
finished in about 300% of the time they normally take.
That at least stopped things from permanently clogging
the test queues.

[Dan Thurman]

On 06/04/2010 07:53 PM, Jon Stanley wrote:
>  On Wed, Jun 2, 2010 at 2:16 PM, Daniel B. Thurman
> <dant(a)cdkkt.com&gt; wrote:
>
>    
> > For those who do manage mirror lists, the specific mirror
> > in question is: "web-ster.com"
> >     
>  I can't see anything wrong with this mirror at the moment - I'm
>  downloading the KDE live ISO from it just for testing.  It seems a
>  little slow for a mirror that claims 250Mbps of bandwidth (I'm getting
>  500-600KB/sec on a connection that could get 6.5MB/s to a well
>  connected mirror near to me), but that could be distance related or
>  whatever.
>     
Yeah, it's strange.  At times it seems to work fine but at
other times it stops working for awhile.  Maybe they were
going through some sort of a "maintenance downtime" or
something - but that does not explain how the yum hang
remained - holding the connection "hostage" - that is the
part I cannot fathom....

As a previous poster explained - it's probably a router
issue...  in any case - I have blacklisted this site and all
is good - so far...

[Jon Stanley]

On Wed, Jun 2, 2010 at 2:16 PM, Daniel B. Thurman <dant(a)cdkkt.com&gt; wrote:

>  For those who do manage mirror lists, the specific mirror
>  in question is: "web-ster.com" 
I can't see anything wrong with this mirror at the moment - I'm
downloading the KDE live ISO from it just for testing.  It seems a
little slow for a mirror that claims 250Mbps of bandwidth (I'm getting
500-600KB/sec on a connection that could get 6.5MB/s to a well
connected mirror near to me), but that could be distance related or
whatever.

[Sam Varshavchik]

Daniel B. Thurman writes:

>  One would think that yum can detect "hangs" and
>  move on to the next mirror to pick up the slack,
>  but this is not the case, or so I am led to believe... 
It's been my consistent experience that yum waits about 30 seconds or so, 
before giving up and going to the next mirror.

[Dan Thurman]

On 06/02/2010 03:20 PM, Sam Varshavchik wrote:
>  Daniel B. Thurman writes:
>
> > One would think that yum can detect "hangs" and
> > move on to the next mirror to pick up the slack,
> > but this is not the case, or so I am led to believe...
>
>  It's been my consistent experience that yum waits about 30 seconds or
>  so, before giving up and going to the next mirror.
>  
Yes, that is true for a "behaving mirror", but in my case,
and on F12, it hangs a very long time (I went away for 30
minutes... came back, and no go!)

So... it was strange.

Seems that the hang in that mirror in question had
a firm grip on the connection and would not time out
at all OR yum thinks somehow the connection is still
live even though there is no activity whatsoever.  I
had to ^C the yum connection to break yum, blacklist
the offending mirror and try again. This time it worked.

Strange.

[Tom Horsley]

On Wed, 02 Jun 2010 14:56:01 -0700
Daniel B. Thurman wrote:

>  > Often it isn't actually the mirror, it is some router
>  > on the path from the mirror to your computer.
>  >     
>  You mean a router outside my DSL router? 
Right, somewhere on the list you get if you do a traceroute.
I had a problem with liberty.edu mirrors for a while
which no one else seemed to be having. Gave up trying to
figure out why, and just added the mirror to the exclude
list in the fastest mirror plugin. I have since
regenned the system and didn't put back the exclusion,
and haven't had any problems, so sometimes it comes
and goes. This networking stuff is all very mysterious :-).

[Frank Murphy]

On 02/06/10 22:56, Daniel B. Thurman wrote:
>  On 06/02/2010 01:46 PM, Tom Horsley wrote:
> > On Wed, 02 Jun 2010 11:16:11 -0700
> > Daniel B. Thurman wrote:
> >
> >   
> >> but then again,
> >> why isn't it that these bad mirrors are being removed
> >> for misbehaving?
> >>     
> > Often it isn't actually the mirror, it is some router
> > on the path from the mirror to your computer.
> >   
>  You mean a router outside my DSL router?
>   
Yep, somewhere between the mirror and your house.
Do a hop count.

-- 
Regards,

Frank Murphy
UTF_8 Encoded
Friend of Fedora

[Dan Thurman]

On 06/02/2010 01:46 PM, Tom Horsley wrote:
>  On Wed, 02 Jun 2010 11:16:11 -0700
>  Daniel B. Thurman wrote:
>
>    
> > but then again,
> > why isn't it that these bad mirrors are being removed
> > for misbehaving?
> >     
>  Often it isn't actually the mirror, it is some router
>  on the path from the mirror to your computer.
>     
You mean a router outside my DSL router?

[Tom Horsley]

On Wed, 02 Jun 2010 11:16:11 -0700
Daniel B. Thurman wrote:

>  but then again,
>  why isn't it that these bad mirrors are being removed
>  for misbehaving? 
Often it isn't actually the mirror, it is some router
on the path from the mirror to your computer.