Hi Phil -

Just wanted to let you know - I reverted back to DB 4.6, recompiled everything, and it all starts up without any issues... there definitely looks like there are some changes with DB 4.7 that makes it incompatible with the current version of PVFS2...

if I get a chance, I'll see if I can probe a bit into the code and see what I can see.

Thanks again for your help and suggestions.

-Mark

Phil Carns wrote:
Oh, now that I think about it a little more, the first server is the only one that actually has file system objects already stored in it initially. It has two objects to keep up with the root "/" directory, and two objects to keep up with the "lost+found" directory. So probably on the first startup it is exercising more db routines than the other servers are.

That doesn't help solve the problem, but it might explain why the first server might have problems on startup while the others do not.

-Phil

Phil Carns wrote:
Ok, let us know what you find out. I'll try 4.7 on my end too, but I have a feeling you'll get to try 4.6 before I get around to trying 4.7 :)

-Phil

Mark J. Hoy wrote:
Thanks Phil -

I'll try installing db 4.6 (or earlier) this afternoon and see if that makes any difference (we don't need 4.7 for anything on our system at the moment, so switching back won't be any issue)...

As for ldd - on all nodes, I'm seeing the same versions:
libdb-4.7.so => /usr0/BerkeleyDB.4.7/lib/libdb-4.7.so (0x00002b3225245000)
       libpthread.so.0 => /lib64/libpthread.so.0 (0x00000030e3300000)
       librt.so.1 => /lib64/librt.so.1 (0x00000030e9000000)
       libc.so.6 => /lib64/libc.so.6 (0x00000030e2c00000)
       /lib64/ld-linux-x86-64.so.2 (0x00000030e2000000)

Thanks again!

-Mark

Phil Carns wrote:
Hi Mark,

Ok, maybe this is some berkeley db compatibility issue; it looks like 4.7 is hot off the presses a couple of weeks ago. Maybe there is something new we need to figure out.

It is strange that you would only have trouble with one particular server, though. Could you double check with "ldd ./pvfs2-server" that they are all linked to the same db library?

-Phil

Mark J. Hoy wrote:
Thanks Phil -

Retrying the re-initialization of the storage space (-f) - says it works, but also provides the additional messages: [S 07/01 10:54] PVFS2 Server on node boston19 version 2.7.1 starting... [E 07/01 10:54] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 07/01 10:54] error in dspace create (db_p->get failed).
[E 07/01 10:54] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 07/01 10:54] error in dspace create (db_p->get failed).
[E 07/01 10:54] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 07/01 10:54] error in dspace create (db_p->get failed).
[E 07/01 10:54] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 07/01 10:54] error in dspace create (db_p->get failed).
[D 07/01 10:54] PVFS2 Server: storage space created. Exiting.

This only happens on the very first server (server1 in my configuration below) - also, I'm using a standard make/make install of BerkeleyDB.4.7 (no odd configuration options other than the --prefix setting to install to a different path)...

Running showcoll on the first server yields:
[E 10:58:15.614847] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 10:58:15.614940] TROVE:DBPF:Berkeley DB: DB->get: Invalid argument
0x00100000 (dspace_getattr output: type = unknown, b_size = 5421632) [E 10:58:15.614990] TROVE:DBPF:Berkeley DB: DB_THREAD mandates memory allocation flag on key DBT
[E 10:58:15.615003] TROVE:DBPF:Berkeley DB: DB->get: Invalid argument
0x00000000 (dspace_getattr output: type = unknown, b_size = 5421632)

Running showcoll (with the same parameters) on any of the other 5 nodes yields: [E 11:01:34.794473] src/io/trove/trove-dbpf/dbpf-mgmt.c line 515: dbpf_collection_geteattr: DB_NOTFOUND: No matching key/data pair found [E 11:01:34.794792] [bt] bin/pvfs2-showcoll(dbpf_collection_geteattr+0x103) [0x414063] [E 11:01:34.794811] [bt] bin/pvfs2-showcoll(main+0x417) [0x4070b7] [E 11:01:34.794822] [bt] /lib64/libc.so.6(__libc_start_main+0xf4) [0x30e2c1d084] [E 11:01:34.794834] [bt] bin/pvfs2-showcoll(aio_fsync64+0x39) [0x406669] Storage space /usr2/pvfs-storage, collection pvfs2-fs (coll_id = 1375400306, *** no root_handle found ***):

... not sure what I'm missing here... Thanks!

-Mark


Phil Carns wrote:
Hello,

I haven't seen that error message in that particular context before. In general, though, it happens on startup when the server finds that it has a handle (storage object) in its directory that doesn't match the ranges specified in the configuration file.

In this specific case it thinks there is a handle with value 0 in your storage space, which shouldn't happen.

Has the server ever started successfully, or is this the first attempt to get it running?

You may want to try just deleting the storage space (the /usr2/pvfs-storage directory) and redoing the "-f" step, if you haven't already.

You could also try running this command:

pvfs2-showcoll -s /usr2/pvfs-storage -c pvfs2-fs

That should list all of the handles in the storage space so that we can confirm if there is really bad data in there or if there is something wrong with the server's startup.

-Phil

Mark J. Hoy wrote:
Hi -

I'm trying to get PVFS2 version 2.7.1 (latest stable) up and running - It compiles correctly without issue and to initialize my storage (via "pvfs2-server -f /path/to/config/file" ) - but I'm having a problem getting the server to start...

every time I try running "sbin/pvfs2-server /path/to/config/file" (where /path/to/config/file is my configuration file generated via pvfs2-genconfig), I keep getting an error: Error: handle 0 is invalid (out of bounds)

The relevant pieces of the log are shown below:
[D 06/27 13:32] PVFS2 Server version 2.7.1 starting.
[E 06/27 13:32] Error: handle 0 is invalid (out of bounds)
[E 06/27 13:32] Error adding handle range 3-1317624576693539402,2635249153387078803-3952873730080618202 to filesystem pvfs2-fs [E 06/27 13:32] Error: Could not initialize server interfaces; aborting.
[E 06/27 13:32] Error: Could not initialize server; aborting.

This seems to happen both using a single-machine configuration, and during a cluster configuration (with 6 machines) - _but_ in the multiple machine configuration, it only happens when I try and start the first I/O node - the other 5 machines startup without issue.

Has anyone else experienced this sort of problem? I've attached a copy of my configuration file below (but changed the machine names to protect the innocent). Also, I'm running on a homogeneous configuration where all six of my machines are running Fedora Core 5, kernel: Linux version 2.6.19.1-001-K8, Dual-Core AMD Opteron(tm) Processor (model 1218, 2.6 GHz), 4GB RAM, and 400 GB of storage on the volume for pvfs2

<Defaults>
       UnexpectedRequests 50
       EventLogging none
       LogStamp datetime
       BMIModules bmi_tcp
       FlowModules flowproto_multiqueue
       PerfUpdateInterval 1000
       ServerJobBMITimeoutSecs 30
       ServerJobFlowTimeoutSecs 30
       ClientJobBMITimeoutSecs 300
       ClientJobFlowTimeoutSecs 300
       ClientRetryLimit 5
       ClientRetryDelayMilliSecs 2000

       StorageSpace /usr2/pvfs-storage
       LogFile /tmp/pvfs2-server.log
</Defaults>
<Aliases>
       Alias server1 tcp://server1:3334
       Alias server2 tcp://server2:3334
       Alias server3 tcp://server3:3334
       Alias server4 tcp://server4:3334
       Alias server5 tcp://server5:3334
       Alias server6 tcp://server6:3334
</Aliases>
<Filesystem>
       Name pvfs2-fs
       ID 1375400306
       RootHandle 1048576
       <MetaHandleRanges>
               Range server1 3-1152921504606846977
Range server6 1152921504606846978-2305843009213693952
       </MetaHandleRanges>
       <DataHandleRanges>
Range server1 2305843009213693953-3458764513820540927 Range server2 3458764513820540928-4611686018427387902 Range server3 4611686018427387903-5764607523034234877 Range server4 5764607523034234878-6917529027641081852 Range server5 6917529027641081853-8070450532247928827 Range server6 8070450532247928828-9223372036854775802
       </DataHandleRanges>
       <StorageHints>
               TroveSyncMeta yes
               TroveSyncData no
       </StorageHints>
</Filesystem>

Thanks!


_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users









_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users



_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

Reply via email to