IssueDcacheRead < CmsTier3

Tags: view all tags
<!-- keep this as a security measure:
   #uncomment if the subject should only be modifiable by the listed groups 
   # * Set ALLOWTOPICCHANGE = Main.TWikiAdminGroup,Main.CMSAdminGroup
   # * Set ALLOWTOPICRENAME = Main.TWikiAdminGroup,Main.CMSAdminGroup
   #uncomment this if you want the page only be viewable by the listed groups
   # * Set ALLOWTOPICVIEW = Main.TWikiAdminGroup,Main.CMSAdminGroup
-->

---+!! %TOPIC%

%TOC%

---++ Symptoms
Summary: %FORMFIELD{"Symptom summary"}%

---++ Occurrences
At what times did this problem occur (used to estimate frequency):

| 2009-05-07 |

---++ Observations
<!--
   #collect here the information which may help to better understand the state of the system or services, e.g.
   #log excerpts, strace output, etc.
   #this also may help to identify the problem if similar conditions arise again
-->

This happened on a *dcache production-1-8-0-15p11(1.17)* installation.

A file cannot be read by gridftp (and therefore also by SRM). The command to read it just hangs forever, the target file on local disk remains at 0 file size.

<pre>
globus-url-copy -v gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/trivcat/store/user/pnef/fastsim_files/CH_V2.7_set5_pp_ttbarw+_jet_soup_5000_1_def_scale_fastsim.root file:///tmp/derek1

Source: gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/trivcat/store/user/pnef/fastsim_files/
Dest:   file:///tmp/
  CH_V2.7_set5_pp_ttbarw+_jet_soup_5000_1_def_scale_fastsim.root  ->  derek1
(hangs forever)
</pre>

Checking whether the file exists

<pre>
echo /pnfs/psi.ch/cms/trivcat/store/user/pnef/fastsim_files/CH_V2.7_set5_pp_ttbarw+_jet_soup_5000_1_def_scale_fastsim.root | dc_get_ID_from_pnfsnamelist.sh |dc_get_cacheinfo_from_IDlist.sh

0002000000000000002ABBD0 t3fs02_cms
</pre>

On the fileserver, the file can be listed:


---+++ Checking whether this is a problem of this particular pool

Does writing to the pool work?

Creating of a new file on that particular pool
<pre>
globus-url-copy file:/tmp/derek2 gsiftp://t3fs02.psi.ch:2811//pnfs/psi.ch/cms/testing/derek20090507a
# where is the file located?
echo /pnfs/psi.ch/cms/testing/derek20090507a |dc_get_ID_from_pnfsnamelist.sh |dc_get_cacheinfo_from_IDlist.sh
0002000000000000002C8B10 t3fs04_cms_1

# we need to move it to the affected pool t3fs02_cms
echo 0002000000000000002C8B10 > IDlist
dc_ppcopy_files.sh t3fs04_cms_1 t3fs02_cms IDlist
dc_get_cacheinfo_from_IDlist.sh IDlist
0002000000000000002C8B10 t3fs02_cms,t3fs04_cms_1

# note: trying to read the file now worked, because it can get it from t3fs04_cms_1

# now we remove it from t3fs04_cms_1
dc_rep_rm_list.sh t3fs04_cms_1 IDlist
dc_get_cacheinfo_from_IDlist.sh IDlist
0002000000000000002C8B10 t3fs02_cms
</pre>

Now we test whether the new file can be read from the pool
<pre>
globus-url-copy -v gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/testing/derek20090507a file:///tmp/derek7
Source: gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/testing/
Dest:   file:///tmp/
  derek20090507a  ->  derek7
(hangs forever)
</pre>

There is one log line in the log of the gridftp door:
<pre>
07 May 2009 12:00:06 Socket OPEN (ACCEPT) remote = /193.40.150.123:34639 local = /192.33.123.42:2811
</pre>

Other files on this pool all show the same behavior.


Trying to transfer a file from the second pool on this machine also ended up with a hanging transfer! So, the condition not specific to a single pool cell.

I was able to trigger a p2p copy of a file to another pool. Regrettably I did not check whether a dcap read on the pool works.

---+++ Logfiles

The only unexplained error I see in =t3fs02Domain.log=, happened two days ago. Unclear whether it is connected.
 
<pre %FILESTYLE%>
06 May 2009 19:25:52 Socket OPEN remote = t3wn05.psi.ch/192.33.123.85:47355 local = /192.33.123.42:36669
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : Exception in runIO for : 0002000000000000002B3288 CacheException(rc=666;msg=Checksum error client=1:c2e0dfc3;file=1:82d23fea)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : CacheException(rc=666;msg=Checksum error client=1:c2e0dfc3;file=1:82d23fea)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.pools.ChecksumModuleV1.setMoverChecksums(ChecksumModuleV1.java:120)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.pools.MultiProtocolPoolV3$RepositoryIoHandler.run(MultiProtocolPoolV3.java:1705)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.util.SimpleJobScheduler$SJob.run(SimpleJobScheduler.java:109)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:885)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:907)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.lang.Thread.run(Thread.java:619)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : Storing incomplete file : 0002000000000000002B3288 with 1476960
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : Stacked Exception (Original) for 0002000000000000002B3288 : CacheException(rc=666;msg=Checksum error client=1:c2e0dfc3;file=1:82d23fea)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : Stacked Throwable (Resulting) for 0002000000000000002B3288 : CacheException(rc=666;msg=Checksum error client=1:c2e0dfc3;file=1:82d23fea)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) : CacheException(rc=666;msg=Checksum error client=1:c2e0dfc3;file=1:82d23fea)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.pools.ChecksumModuleV1.setMoverChecksums(ChecksumModuleV1.java:120)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.pools.MultiProtocolPoolV3$RepositoryIoHandler.run(MultiProtocolPoolV3.java:1912)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at diskCacheV111.util.SimpleJobScheduler$SJob.run(SimpleJobScheduler.java:109)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:885)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:907)
05/06 19:26:52 Cell(t3fs02_cms@t3fs02Domain) :  at java.lang.Thread.run(Thread.java:619)
06 May 2009 19:27:22 Socket CLOSE remote = t3wn05.psi.ch/192.33.123.85:47355 local = /192.33.123.42:36669
06 May 2009 19:27:44 remove entry for: 0002000000000000002B3288
</pre>


In =/var/adm/messages.0= there are signs of a disk problem!
<pre %FILESTYLE%>
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.warning] WARNING: /pci@1,0/pci1022,7458@4/pci11ab,11ab@1/disk@3,0 (sd30):
May  5 06:07:55 t3fs02  Error for Command: write(10)               Error Level: Retryable
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Requested Block: 326265585                 Error Block: 326265585
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Vendor: ATA                                Serial Number:
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Sense Key: Aborted_Command
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU: 0x0
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.warning] WARNING: /pci@1,0/pci1022,7458@4/pci11ab,11ab@1/disk@3,0 (sd30):
May  5 06:07:55 t3fs02  Error for Command: write(10)               Error Level: Retryable
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Requested Block: 326265329                 Error Block: 326265329
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Vendor: ATA                                Serial Number:
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    Sense Key: Aborted_Command
May  5 06:07:55 t3fs02 scsi: [ID 107833 kern.notice]    ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU: 0x0
...
...
Apr 27 18:01:12 t3fs02 scsi: [ID 107833 kern.warning] WARNING: /pci@1,0/pci1022,7458@3/pci11ab,11ab@1/disk@1,0 (sd20):
Apr 27 18:01:12 t3fs02  Error for Command: read(10)                Error Level: Retryable
Apr 27 18:01:12 t3fs02 scsi: [ID 107833 kern.notice]    Requested Block: 380806823                 Error Block: 380806823
Apr 27 18:01:12 t3fs02 scsi: [ID 107833 kern.notice]    Vendor: ATA                                Serial Number:
Apr 27 18:01:12 t3fs02 scsi: [ID 107833 kern.notice]    Sense Key: Aborted_Command
Apr 27 18:01:12 t3fs02 scsi: [ID 107833 kern.notice]    ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU: 0x0
...
...
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.warning] WARNING: /pci@2,0/pci1022,7458@8/pci11ab,11ab@1/disk@4,0 (sd47):
Apr  1 07:17:02 t3fs02  Error for Command: write(10)               Error Level: Retryable
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Requested Block: 239734606                 Error Block: 239734606
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Vendor: ATA                                Serial Number:
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Sense Key: Aborted_Command
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU: 0x0
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.warning] WARNING: /pci@2,0/pci1022,7458@8/pci11ab,11ab@1/disk@4,0 (sd47):
Apr  1 07:17:02 t3fs02  Error for Command: write(10)               Error Level: Retryable
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Requested Block: 239734862                 Error Block: 239734862
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Vendor: ATA                                Serial Number:
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    Sense Key: Aborted_Command
Apr  1 07:17:02 t3fs02 scsi: [ID 107833 kern.notice]    ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU: 0x0
</pre>



---++ Solution or Workaround

---+++ Pools need to be restarted

Set the pools to read only:
<pre>
dc_set_pools_readonly.sh cms_broken_pools.lst
</pre>

Restart the pools on the affected node t3fs02 (note that a disabled pool will be enabled after restart!):
<pre>
/opt/d-cache/bin/dcache stop pool
/opt/d-cache/bin/dcache start pool
</pre>


Test whether a file can be read by gridftp:
<pre>
globus-url-copy -v gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/testing/derek20090507a file:///tmp/derek103
Source: gsiftp://t3fs02.psi.ch:2811///pnfs/psi.ch/cms/testing/
Dest:   file:///tmp/
  derek20090507a  ->  derek103

ls -l /tmp/derek103
-rw-r--r--  1 feichtinger cms 51200 May  7 16:00 /tmp/derek103
</pre>

*OK*

Set the pools to read/write, again.
<pre>
dc_set_pools_readonly.sh -n cms_broken_pools.lst
</pre>

---++ Monitoring for this condition
<!--
   #how can this condition be recognized automatically, if at all?
-->


-- Main.DerekFeichtinger - 07 May 2009
IssueForm
Affected Service	gsiftp
Symptom summary	gridftp (and srmcp) file read hangs forever
Reason Understood	no
Solution Exists	workaround
Topic revision: r3 - 2009-05-07 - DerekFeichtinger
CmsTier3
User Pages
Main Page
Policies
Physics Groups
Steering Board Meetings
Admin Pages
AdminArea
Cluster Specs