sks-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sks-devel] Constant high iowait. Infinte recon loop? What to do?


From: Andrey Korobkov
Subject: Re: [Sks-devel] Constant high iowait. Infinte recon loop? What to do?
Date: Sun, 26 Jun 2011 00:05:53 +0400

On Thu, 16 Jun 2011 12:02:19 +0000
Kim Minh Kaplan <address@hidden> wrote:

> Andrey Korobkov:
> 
> > I can't understand, what's with my server:
> > It's iowait is so high, that it responds so slow, that keyserver pool 
> > treats it as unresponding.
> > Why does it requesting many hundreds and thousands of keys, and what does 
> > it really doing with them?
> 
>     2011-04-11 23:54:32 Applying 200 changes
>     [...]
>     2011-04-11 23:55:24 add_keys_merge failed: Eventloop.SigAlarm
>     2011-04-11 23:55:25 Key addition failed: Eventloop.SigAlarm
> 
> For some reason your server does not seem to be able to update its
> database in the allocated time for a request. You could try to lower
> http_fetch_size in your sksconf. It defaults to 100; try 50 or even 1 if
> needed (and be ready for a looong catch up time).
> 
Thanks for the advice. It helps for some time.
But later I just dicovered, that sks database had grown to the whole disk space!
(Why? May be, it was DB_LOG_AUTOREMOVE flag that was needed?)
I've recovered DB from fresh dump and applied custom DB_CONFIG (thanks to Jeff 
Johnson :) ).
I hope, the keyserver should run well since now...

> Is your SKS server running on some form of low performance server?
keyserver.fryxell.ru is running on my own home 24*7 Parabola GNU/Linux-libre 
machine
(dedicated physical server! :) ) The server hardware is rather old, but still 
powerful:
AMD Athlon XP 2600+, RAM: 1287368 kB, HDD: (160 + 20) GB IDE.
Connectivity: symmetrical 10 Mbit/s FTTB.
Seems to be enough for a keyserver (many VPS-es have much less resources), 
doesn't it?

Rather it may be me, who don't know how to configure keyserver's database 
well... :)

P.S. Also, some problems may have appeared due to NAT-ing my machine for some 
days?
Can SKS recon it's DB behind the NAT? (having ports forwarded to it, but 
listening on
private addresses 192.168.1.0/24? May be, the recon protocol announces that 
private address for HKP too?)
In either case, the machine isn't NAT-ed now, so SKS should run well.
I'm just curious, whether NAT can cause such a trouble? If so, I suggest adding 
a warning to the FAQ...

-- 
Andrey Korobkov

Attachment: pgp4IfFnw0jsb.pgp
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]