Bug 36285

Summary:	Xinet daemon is very slow
Product:	[Retired] Red Hat Linux	Reporter:	Bill Farrell <billfarr>
Component:	xinetd	Assignee:	Trond Eivind Glomsrxd <teg>
Status:	CLOSED NOTABUG	QA Contact:	David Lawrence <dkl>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	7.0
Target Milestone:	---
Target Release:	---
Hardware:	i386
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2001-04-17 16:13:53 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Bill Farrell 2001-04-17 16:13:49 UTC

From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows 98)


> I'm wondering about xinetd for RH7.  I just recently bought and installed
> RH7 (the Pro version) and installed it on a machine that was previously
> running RH6.2.
>
> After a wee while with the editor rearranging paths (since /etc now has a
> lot of configs in subdirectories) it looked great and I thought I was 
about
> ready to put it into service.  I really like the new organisation of
> configs...being in subdirectories, one can easily document-as-you-go.
>
> Then I noticed something weird.  I had two machines, one next to the 
other
> and even next to one another in the network hub.  From the RH6.2 machine 
to
> the RH7 machine, pings were running about the expected 0.5ms mark.  Pings
> from RH7 to the RH6.2 machine were running around 300-500ms.  Telnet also
> took an exhorbitant time to connect and performed lackadaisically, even 
with
> one user connected (me).
>
> Of course, I'm not at all familiar with the new xinetd and wonder if 
there's
> a tuning issue involved.
>
> All 20-some of my servers are using a DEC RLT8139-based NIC (which RH 
seems
> to love).  I wound up taking that particular server BACK to 6.2 where it
> performs up-to-snuff.  All the network-related problems disappeared.
>
> I haven't seen any benches or mentions of xinetd performance as opposed 
to
> inetd performance.  RH7 stays on the shelf until I can figure out why 
xinetd
> performs so awfully.
>


Reproducible: Always
Steps to Reproduce:
1. Set up a 6.2 machine beside a 7.0 machine
2. From the 6.2 machine, ping the 7.0 machine.  Note times.
3. From the 7.0 machine, ping the 6.2 machine.  Note times.
4. Try loading heavy web pages from each.  Make a nasty one with lots of 
graphics and crap.  Note which loads faster (6.2 every time).
	

Actual Results:  The 6.2 machine served faster and more reliably.  The 7.0 
machine tended to hang under heavy loads.

Expected Results:  7.0 should have served identically or better.

Please feel free to contact me and I'll be happy to go re-create the 
scenario.

The machine composition upon which I was running 7.0 was:
AMD K6/3 800MHz CPU, 196M RAM, 60G of IDE disk (2x30G), ATI RAGE-II video 
(no desktop, bare server).

Comment 1 Trond Eivind Glomsrxd 2001-04-17 18:11:05 UTC

1) 7.1 should have better performance than 7 out-of-the-box
   One important change is when in the process access control is implemented
2) remove /etc/hosts.{allow,deny}, use "only_from" inside the xinetd config files
   or even better - firewalling