Hya,

Set up
Debian testing, (buster amd64) bibus, using sqlite, database file on NFS dirve.

Situation
After dist-upgrade, bibus stopped working. It looks like that an error is locked database, according to sqlitebrowser.

Troubleshooting
This thread says some trouble with database file on NFS drive. I tried renaming and several tricks, without success. No other processes use same database file. lsof and fuser fails to detect any clue.
Whole system reboot and fsck did not solve situation.
I found that item #5 of this thread and this one say that some trouble with locking process.

I guess if I am on a right track, it is far beyond of whatever I can do. Log files say that many python files are updated, but neither nfs nor kernel related ones. (bibus is a python2 application.)

For the time being, I can move database file to local drive, although for data security reason, it is not feasible.

Question
Has anybody ever run into similar situation? Is there any other workaround?

Closing remark
Thank you in advance

Hya

I also found that libreoffice files on nfs are also affected. App says that "Document file @#.od? is locked for editing by Unknown user"

There is no .lock file and so far all of od? files on nfs drive are affected. Once file is copied to local drive, it can be edited. I found bug #884284, but it does not look related.

I will update.

cheers

++ added
I recompiled kernel (4.18.10-rt5), but did not solve. System is up-to-date as of 19/Oct.
Following applications can edit files on nfs drive. - pycharm, vi, gimp

Hya

update: (nfs is v3)

Still in troubleshooting.
With some other nfs server, no trouble.
Kernel option about nfsv3 ACL protocol extension is NOT related.

Found error log!
/var/log says
server kernel: lockd: cannot monitor client <- this line shows up when libreoffice chokes

Related web pages: 78004 emanugrum 23373

Still I have not found solution yet.

cheers

Hya

Update:

I am not sure whether this is related, but on NON working nfs server node,The user '_rpc' is running rpcbind.

Neither re-configure nor re-install of rpcbind package works.

+++ added on edit
Probably, the info above is irrelevant. It is a new feature introduced recently.

cheers

Hya,

I guess my previous post (#4) is totally off track.

Now, I know that
/proc/sys/sunrpc is missing on nfs server. (rpcbind is running!)
/proc/sys/sunrpc exists on nfs client.
recent upgrade of rpcbind (0.2.3-0.6 -> 1.2.5-0.3) is NOT related.

Debugging info on client says that "lockd: server returns status 2"

Currently, I cannot go further, because I cannot set debug info on server. (This info does not help.)

With the working nfs server, it is possible to set debug info. For some obvious reason, no error (on both client and server).

cheers

++ added
CONFIG_SRNRPC_DEBUG needs to be on to get /proc/sys/sunrpc

However, still problem persists. Problematic log (/var/log/messages) is

Hya

correction.

Probably recent upgrade of rpcbind IS related. (post #5 is incorrect)

When I posted #5, I only rebooted nfs host. (host was upgraded and rebooted. client was 0.2.3-0.6)
After reboot (or whatever reason), all nfs hosts (1.2.5-0.3) are dead, regardless of client version.

cheers

+ added
Same error happens with sqlite3.
Is this a bug?

Hya,

Update

Recent dist-upgrade (plus new kernel) did not solve this situation.

Found this thread., and followed post #1. on nfs host node works!! For some unknown reason, rpc.statd does not kick in.

I am not sure if this is same nature as bug #805167

Not solved, but one step ahead.

cheers

It looks like that some rule was overwritten on system upgrade.

Solution may be in this thread.

cheers