Hi all,
We're currently seeing similar TLS issues in osg-test with xrdcp. It's
why we haven't released 5.6.x in the OSG repos yet:
https://opensciencegrid.atlassian.net/browse/SOFTWARE-5623?focusedCommentId=380033.
I'm also CC'ing Lincoln who is seeing this issue out in the wild with a
5.6.1 client against a 5.5.5 server (see attached). For comparison, I've
also attached a successful copy for a 5.5.5 client vs 5.5.5 server for
the same file.
Bockjoo: would you be able to try downgrading to a 5.5 client and seeing
if that resolves your issue?
Thanks,
Brian
On 8/28/23 08:42, Bockjoo Kim wrote:
> Hi Guilherme,
>
> The client machine (xrootd 5.6.1) that I had the issue is a fully
> credible OSG CE.
>
> So, there should be no issue with the CA and X509_CERT_DIR.
>
> I have used this python script to reproduce the issue on the client
> machine:
>
> #############################################
>
> import os
> import sys
> import errno
> import subprocess
> import zlib
> import random
> from XRootD import client
> from XRootD.client.flags import OpenFlags
>
> ENDPOINT='cmsio2.rc.ufl.edu:1094'
> SAM_TEST_FILE='/store/mc/SAM/GenericTTbar/AODSIM/CMSSW_9_2_6_91X_mcRun1_realistic_v2-v1/00000/A64CCCF2-5C76-E711-B359-0CC47A78A3F8.root'
>
>
> print ("XRootD Client Versin",client.__version__)
> cmd = [ "xrdfs "+ENDPOINT+" query config version" ]
> try:
> result = subprocess.run(cmd, shell=True, capture_output=True,
> text=True)
> print("XRootD Server Version", result.stdout)
> except subprocess.TimeoutExpired:
> print("connecting to endpoint timed out")
>
>
> os.environ["X509_CERT_DIR"] =
> "/cvmfs/cms.cern.ch/grid/etc/grid-security/certificates"
> os.environ["X509_USER_PROXY"] = "/home/bockjoo/.cmsuser.proxy"
> os.environ["X509_USER_PROXY_NONCMS"] = "/home/bockjoo/.griduser.proxy"
> os.environ["XRD_NETWORKSTACK"] = "IPv4"
> with client.File() as f:
> status, response = f.open("root://" + ENDPOINT + "/" + \
> SAM_TEST_FILE, flags=OpenFlags.READ, timeout=90)
> if ( not status.ok ):
> print (("\nopen(root://%s/%s, flags=OpenFlags.READ,
> time" + \
> "out=90)\nXRootDStatus.code=%d \"%s\"\n") % \
> (ENDPOINT, SAM_TEST_FILE, status.code, \
> status.message.replace("\n", "")))
> pass
> status, data = f.read(offset=0, size=65536, timeout=90)
> if ( not status.ok ):
> print(("\n%s\nread(offset=0, size=65536,
> timeout=90)\n" + \
> "XRootDStatus.code=%d \"%s\"\n") %
> (SAM_TEST_FILE, \
> status.code, status.message.replace("\n", "")))
> pass
> print ("Open Status",status.ok)
>
> #############################################
>
> You can choose the endpoint and the file of your choosing with the
> 5.5.5 server
>
> to test it.
>
> Thanks,
>
> Bockjoo
>
> On 8/28/23 09:33, Guilherme Amadio wrote:
>> Dear Bockjoo,
>>
>> On Sat, Aug 26, 2023 at 04:14:28PM -0400, Bockjoo Kim wrote:
>>> Hi,
>>>
>>> I am seeing a python XRootD file open issue for the 5.6.1 client with a
>>> 5.5.5 server :
>>>
>>> =============================================================
>>>
>>> XRootD Client Versin 5.6.1
>>> XRootD Server Version v5.5.5
>>>
>>> open(root://cmsio2.rc.ufl.edu:1094//store/mc/SAM/GenericTTbar/AODSIM/CMSSW_9_2_6_91X_mcRun1_realistic_v2-v1/00000/A64CCCF2-5C76-E711-B359-0CC47A78A3F8.root,
>>>
>>> flags=OpenFlags.READ, timeout=90)
>>> XRootDStatus.code=110 "[FATAL] TLS error: resource temporarily
>>> unavailable: Unable to connect to cmsio2.rc.ufl.edu; error_ssl"
>>>
>>> ---------------------------------------------------------------------------
>>>
>>> ValueError Traceback (most recent
>>> call last)
>>> /tmp/ipykernel_4179061/812350213.py in <module>
>>> 40 status.message.replace("\n", "")))
>>> 41 #pass
>>> ---> 42 status, data = f.read(offset=0, size=65536,
>>> timeout=90)
>>> 43 if ( not status.ok ):
>>> 44 print(("\n%s\nread(offset=0, size=65536,
>>> timeout=90)\n" + \
>>>
>>> /opt/cms/services/anaconda3/lib/python3.9/site-packages/XRootD/client/file.py
>>>
>>> in read(self, offset, size, timeout, callback)
>>> 124 return XRootDStatus(self.__file.read(offset, size,
>>> timeout, callback))
>>> 125
>>> --> 126 status, response = self.__file.read(offset, size, timeout)
>>> 127 return XRootDStatus(status), response
>>> 128
>>>
>>> ValueError: I/O operation on closed file
>>>
>>> ===============================================================
>>>
>>> Here, XRootD Server is configured with TLS.
>>>
>>> If I remove TLS configuration of the 5.5.5 server, there is no issue.
>>>
>>> If I switch to 5.5.5 client, there is no issue.
>>>
>>> Is this expected?
>> It may or may not be. When I wrote the patch, I tested several scenarios
>> (see
>> https://github.com/xrootd/xrootd/pull/2031#issuecomment-1589380486).
>> The error message that you see is likely caused by a client that cannot
>> validate the server with TLS (because it does not have the proper CA
>> certificates installed locally). So I suggest you to try with xrdcp
>> --notlsok option, or export X509_CERT_DIR=/dev/null to force the client
>> into not being able to do TLS at all. If the directory
>> /etc/grid-security
>> exists on your machine, but the client cannot verify the server, and TLS
>> is enforced, then this behavior is expected. Otherwise, please export
>> XRD_LOGLEVEL=Dump, re-run the command and send us the output so I can
>> investigate this issue further. You may also want to install the proper
>> certificates to let the client validate the server to be able to use
>> TLS.
>>
>> Best regards,
>> -Guilherme
>
> ########################################################################
> 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
