Hi Andy,
I am still investigating using the proxy cache for read only instance but Sphenix wanted to be able to write also from the production system. It is next on my list of work to do.
Thank you for the update.
Regards,
Doug
-----Original Message-----
From: <[log in to unmask]> on behalf of Andrew Hanushevsky <[log in to unmask]>
Date: Monday, March 21, 2022 at 2:36 PM
To: Doug Benjamin <[log in to unmask]>
Cc: Wei Yang <[log in to unmask]>, xrootd-l <[log in to unmask]>
Subject: Re: how do I increase request size for xrootd
Hi Doug,
Then it would follow it's Lustre's internal transfer size I would think.
A normal xrootd data server will never ask for more bytes than what the
client requested. Now, if you use a disk caching proxy server you get
complete control of that transfer chain. Unfortunately, you can't write.
However, using a memory caching server you can read and write but it's a
bit more restricted because you're limited by the amount of memory you
have.
Anyway, I thought you were going to use a cache with Lustre. What
happened?
Andy
On Mon, 21 Mar 2022, Doug Benjamin wrote:
> 256k requests size comes from looking at the lustre file system internal monitoring as seen at the lustre file servers.
>
> When the code uses the lustre client directly, the requests to the lustre file servers are much larger.
>
> Both Minio and Xrootd are using smaller request sizes to Lustre backend.
>
>
>
> -----Original Message-----
> From: "Yang, Wei" <[log in to unmask]>
> Date: Monday, March 21, 2022 at 10:50 AM
> To: Doug Benjamin <[log in to unmask]>, xrootd-l <[log in to unmask]>
> Cc: "[log in to unmask]" <[log in to unmask]>
> Subject: Re: how do I increase request size for xrootd
>
> Hi Doug,
>
> " Config asynchronous I/O has been disabled!" is expected, as you don't need async IO in LAN with a direct access to storage. Async is useful in WAN and proxy environment.
>
> What is "request sizes of 256k" referring to?
>
> ?--
> Wei?
>
> ________________________________________
> From: Doug Benjamin <[log in to unmask]>
> Sent: Monday, March 21, 2022 8:34 AM
> To: xrootd-l
> Cc: Hanushevsky, Andrew B.; Yang, Wei
> Subject: how do I increase request size for xrootd
>
> Hello,
>
> I am using several xrootd servers as front ends to distributed Lustre filesystem. The 3 servers get data from Lustre client via mounted Lustre FS.
>
> Right now we are seeing request sizes of 256k. Would like to increase it to 1M if possible. Most of the access is read only but we do have a read/write use case.
>
> Also the xrootd data server when it started up I saw this line in the xrootd log -
>
> " Config asynchronous I/O has been disabled!"
>
>
> Here is my configuration file -
>
> all.sitename sPHENIX
> xrd.port 1094
> all.export /sphenix/lustre01/sphnxpro
> all.role server
> all.role manager if xrdsphenix01.rcf.bnl.gov xrdsphenix02.rcf.bnl.gov
> all.manager xrdsphenix01.rcf.bnl.gov:1213
> all.manager xrdsphenix02.rcf.bnl.gov:1213
> cms.allow host xrdsphenix01.rcf.bnl.gov
> cms.allow host xrdsphenix02.rcf.bnl.gov
> cms.allow host dcsphdoor01.rcf.bnl.gov
> cms.allow host dcsphdoor02.rcf.bnl.gov
> cms.allow host sphnxrootd01.sdcc.bnl.gov
> xrootd.seclib libXrdSec.so
> sec.protocol sss -s /etc/xrootd/sss.keytab_sphenix.grp
> ofs.authorize
> acc.audit deny
> acc.authdb /etc/xrootd/auth_file.sphenix
> acc.authrefresh 60
> ofs.authorize 1
> xrd.network keepalive kaparms 10m,1m,5
> xrd.timeout idle 60m
> xrootd.trace request response
> http.trace request response
>
> Thank you in advance.
>
> Regards,
> Doug Benjamin
>
>
>
########################################################################
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
