XROOTD-L Archives

Support use of xrootd by HEP experiments

XROOTD-L@LISTSERV.SLAC.STANFORD.EDU

Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Andrew Hanushevsky <[log in to unmask]>
Reply To:
Support use of xrootd by HEP experiments <[log in to unmask]>
Date:
Wed, 22 Aug 2018 13:33:15 -0700
Content-Type:
TEXT/PLAIN
Parts/Attachments:
TEXT/PLAIN (117 lines)
Hi George,

On Wed, 22 Aug 2018, George Patargias - UKRI STFC wrote:

> Is the directive xrd.port the one I should be using to define the port 
> number for each instance (inside the if block)?
You can have in the confg file gaurded by an if. Some people prefer to 
specify t on he start-up command line. It really depends on your 
philosophy, either will work.

> Does the redirector need to be a seperate machine or it can be DNS 
> alias of all the nodes where both xrootd instances are running? Can I 
> use xrootd.redirect to forward the incoming requests from the standard 
> port 1094 to the ports of the two instances?
The redirector can run on any machine you want here, including the one 
that hosts the two instances, it becomes a third instance. Note that since 
you will run a simple rediretor (i.e. onl xrootd no cmsd) please don't 
specify the "all.role" directive -- that's only meant for installations 
using the cmsd. Yes, the xrootd.redirect specifies the host name (here the 
same one) and the port number.

Andy

  >
> George
>
> ________________________________
> From: Andrew Hanushevsky <[log in to unmask]>
> Sent: 17 August 2018 18:00:34
> To: Patargias, George (STFC,RAL,SC)
> Cc: [log in to unmask]
> Subject: Re: XRootD instances with different config
>
> Hi George,
>
> It is completely possible and the complexity is only driven by how
> transparent you wish this to be. Let first start with running two xrootd
> instances on the same node. The steps:
>
> a) Each instance would be started with a different instnace name (see the
> -n option on the xrootd command line). Of course, the first instance can
> deault to no name which he other would have a name, say "-n alice".
> http://xrootd.org/doc/dev47/xrd_config.htm#_Toc483004720
>
> b) A single configuration file can be used by special casing direcyives
> using the "if" directive. For instance,
>
> if named alice
> <directives specific to the alice instance>
> else
> <directives specific to the non-alice instance>
> fi
> <common directives>
> http://xrootd.org/doc/dev42/Syntax_config.htm#_Toc427321307
>
> c) Each instnace would be assigned its own port number. This can be done
> in the config file or via command line. This is likely the one requirement
> that is most annoying but is one required by all OS's.
>
> Now, say you want to use the same port nuber for both instance. Youcan do
> that by layering both instance with a redirector. The redirector would
> have the standard port number (say 1094) and evryone would connect to it.
> It would then decide in the incomming client needs to go to the un-named
> instance or to the special alice instance and redirect the client to the
> "right" one. Each of hose would simply be running on soe abitrary port.
>
> If the decision can be made based on file path then you need not write any
> plugin but use the xrootd.redirect directive to accomplish this.
> http://xrootd.org/doc/dev48/xrd_config.htm#_Toc496911346
>
> Otherwise, you will need to write a plugin to do more detailed decision
> making. Since, I don't know enough here I can't tell in this note.
> However, feel free to contact me if a plugin is warrented in this case.
>
> Andy
>
> On Fri, 17 Aug 2018, George Patargias - UKRI STFC wrote:
>
>> Hello,
>>
>>
>> Apologies if this issue has been discussed before.
>>
>>
>> This is in relation to the use of xrootd as part of the CASTOR storage service in RAL Tier-1. We are planning to set up a single CASTOR storage instance for all LHC and no-LHC VOs.
>>
>>
>> Right now, there is a xrootd@manager service running on each CASTOR headnode controlled by /etc/xrootd/xrootd-manager.cfg. However, one VO
>>
>> (ALICE in particular) needs certain lines (for authorization) in the xrootd-manager.cfg that other VOs do not need (and don't want).
>>
>>
>> My question. Is it possible to set up two xrootd instances running on each headnode each one with its own xrootd-alice.cfg? I mean one "generic" instance (e.g. xrootd@lhc) for all LHC VOs (including those non-LHC VOs that do not need special authorisation config in xrootd) and one instance (e.g. xrootd@alice) that will by used only by the ALICE VO in which the xrootd-alice.cfg will contain the neccessary ALICE-specific directives.
>>
>>
>> If this is possible what other xrootd config directives (all.export ?) we need so that given a single endpoint (root://castor-stager.gridpp.rl.ac.uk//) the ALICE users will be directed to the xrootd@alice service and all the other users will be directed to the xrootd@lhc service?
>>
>>
>> Many thanks in advance,
>>
>>
>> George
>>
>> ########################################################################
>> Use REPLY-ALL to reply to list
>>
>> To unsubscribe from the XROOTD-L list, click the following link:
>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>>
>

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the XROOTD-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1

ATOM RSS1 RSS2