Instructions for using beegfs-client-compat

[Nathan R.M. Crawford]

Hi All,

  I am trying to use a single BeeGFS-6 client to mount one file system from a v6 server and one from a v2015.03 server. The only documentation I can find is a couple posts in this list from 2014. Here is my best guess at the proper procedure, although it is obviously wrong somewhere as I cannot mount the older FS.

1) Take working v6 config and convert to a multi-mount setup. in /etc/beegfs/newfs.d/ and /etc/beegfs/oldfs.d/ have unique client, helperd, and mount config files with different port offsets.

2) Download and install beegfs-client-compat-2015.03-2015.03.r26-el7.noarch.rpm

3) Build and install module:

  cd /opt/beegfs/src/client/beegfs_client_compat_module_2015.03/build

  make TARGET=beegfs_201503 BEEGFS_OPENTK_IBVERBS=1

  make install TARGET=beegfs_201503

4) modprobe beegfs_201503

At this point the module looks OK, and I can mount the v6 newfs with:

mount -t beegfs beegfs_newfs /newfs -ocfgFile=/etc/beegfs/newfs.d/beegfs-client.conf,_netdev

but trying to mount the old with:

mount -t beegfs_201503 beegfs_oldfs /oldfs -ocfgFile=/etc/beegfs/oldfs.d/beegfs-client.conf,_netdev

fails with this in the syslog:

beegfs_201503: modprobe(130515): File system registered. Type: beegfs_201503. Version: 2015.03-r26

beegfs_201503: mount(130659): External logger (provided by beegfs-helperd) is unreachable => logging currently disabled

beegfs_201503: beegfs_201503_X(130710): Helper daemon is unreachable => corresponding functionality (e.g. hostname resolution) currently unavailable

beegfs_201503: mount(130659): Mount sanity check: Communication with user-space helper daemon failed. Is beegfs-helperd running?

beegfs_201503: mount(130659): Mount sanity check failed. Canceling mount. (Log file may provide additional information. Check can be disabled with sysMountSanityCheckMS=0 in the config file.)

  Since the error may take some time to work through, and could be an unrelated network issue, I am mainly posting to see if steps 1-4 are the right way to do this. Are there official instructions posted somewhere? Backwards compatibility, at least during transitional periods, seems like something useful for a lot of people.

Thanks,

Nate

--

Dr. Nathan Crawford              nathan....@uci.edu
Modeling Facility Director
Department of Chemistry
1102 Natural Sciences II         Office: 2101 Natural Sciences II
University of California, Irvine  Phone: 949-824-4508
Irvine, CA 92697-2025, USA

[Nathan R.M. Crawford]

Hi Bernd,

  Dingdingdingdingding! That seems to allow mounting!

  I will try the autobuild using the scripts next.

Thanks a ton,

Nate 

On Fri, Sep 8, 2017 at 12:24 PM, Bernd Schubert <"Bernd Schubert"@fastmail.fm> wrote:

Hi Nathan,

I don't remember the details anymore and don't have time to try this on
my own now. I would do the following in
/etc/beegfs/oldfs.d/beegfs-client.conf

# use the IP, not the same
sysMgmtdHost = <IP-of-the-mgmtd>

logType = syslog

logHelperdIP = <IP-of-any-system-running-the-old-helperd>


Also, I'm not sure if anything has changed in the mean time, but
actually /etc/init.d/beegfs-client *was* supposed to build the old
module as well (this step just didn't work with older versions that
didn't know about the compat module yet).

Hope it helps,
Bernd

[Toby Darling]

Hi Nathan (and everyone else)

Many thanks for your post, it was most useful as we have two 2015.03
clusters and now a new v7 cluster in test. My clients can now talk to
all 3 thanks to your pointers.

The thing that was going to trip me up was our nagios monitoring system
which uses beegfs-ctl to gather cluster stats. As there's no
beegfs-ctl-compat-2015, I grabbed /opt/beegfs/lib/libbeegfs-opentk.so
and /opt/beegfs/sbin/beegfs-ctl from a 2015.03 installation.

However, beegfs-ctl talks to the kernel client to get it's config from
/proc/fs/beegfs/<ID>/config. That path is hard coded in the client
source. When you build the client-compat, the kernel module is called
beegfs-201503 by default, so the config is under /proc/fs/beegfs-201503/...

You need to edit
/opt/beegfs/src/client/beegfs_client_compat_module_2015.03/source/filesystem/FhgfsOpsIoctl.c

changing line 218

cfgFileStrLen = os_scnprintf(fileName, BEEGFS_IOCTL_CFG_MAX_PATH,
"/proc/fs/beegfs/%s/config",

to resemble

cfgFileStrLen = os_scnprintf(fileName, BEEGFS_IOCTL_CFG_MAX_PATH,
"/proc/fs/beegfs-201503/%s/config",

[tweak as necessary if you're not using the default kernel module name]

Now I can

LD_LIBRARY_PATH=. ./beegfs-ctl ...

and talk to my old clusters, while using /usr/bin/beegfs-ctl to talk to
the new one.

This feels like a hack, if anyone knows of a cleaner way of achieving
this (like a macro in that path?), please let me know.
A win for open source, though. Thanks BeeGFS!

Cheers
Toby

> Dr. Nathan Crawford nathan....@uci.edu <mailto:nathan....@uci.edu>

> Modeling Facility Director
> Department of Chemistry
> 1102 Natural Sciences II Office: 2101 Natural Sciences II
> University of California, Irvine Phone: 949-824-4508
> Irvine, CA 92697-2025, USA
>

> --
> You received this message because you are subscribed to the Google
> Groups "beegfs-user" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to fhgfs-user+...@googlegroups.com
> <mailto:fhgfs-user+...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

Cheers
Toby
--
Toby Darling, Scientific Computing (2N249)
MRC Laboratory of Molecular Biology
Francis Crick Avenue
Cambridge Biomedical Campus
Cambridge CB2 0QH
Phone 01223 267070