NFS service inside a container

Hi folks,

I would like to run an NFServer inside a container. The idea is to mirror the whole NFS service container to a fallback host using drbd, and to provide preconfigured NFS containers to whoever needs it.

Unfortunately the kernel (3.14 of Debian Jessie, backported to Wheezy) did not share my enthusiasm:

Jun 13 15:33:55 srvl011a kernel: [  725.372466] WARNING: CPU: 6 PID: 11134 at /build/linux-v1L7fI/linux-3.14.4/fs/nfsd/nfs4recover.c:1195 nfsd4_umh_cltrack_init+0x42/0x50 [nfsd]()
Jun 13 15:33:55 srvl011a kernel: [  725.372468] NFSD: attempt to initialize umh client tracking in a container!
Jun 13 15:33:55 srvl011a kernel: [  725.376800] Modules linked in: veth sha256_ssse3 sha256_generic hmac drbd lru_cache libcrc32c nfsd auth_rpcgss oid_registry nfs_acl nfs lockd fscache sunrpc bonding bridge stp llc dm_crypt ioatdma radeon iTCO_wdt iTCO_vendor_support i5400_edac lpc_ich mfd_core rng_core ttm edac_core i2c_i801 i5k_amb coretemp drm_kms_helper psmouse shpchp serio_raw joydev pcspkr evdev drm parport_pc parport processor button kvm tpm_tis tpm thermal_sys ext4 crc16 mbcache jbd2 crc32c btrfs xor raid6_pq dm_mod hid_generic usbhid hid sd_mod crct10dif_generic crc_t10dif crct10dif_common sg sr_mod cdrom ata_generic ehci_pci ata_piix uhci_hcd ehci_hcd libata usbcore e1000e usb_common igb i2c_algo_bit i2c_core dca ptp pps_core 3w_9xxx scsi_mod
Jun 13 15:33:55 srvl011a kernel: [  725.376868] CPU: 6 PID: 11134 Comm: rpc.nfsd Tainted: G        W I  3.14-0.bpo.1-amd64 #1 Debian 3.14.4-1~bpo70+1
Jun 13 15:33:55 srvl011a kernel: [  725.376870] Hardware name: Supermicro X7DW3/X7DWN+, BIOS  1.2 11/04/2008
Jun 13 15:33:55 srvl011a kernel: [  725.376872]  0000000000000000 ffffffffa0740758 ffffffff814eb7c7 ffff8810278cbd78
Jun 13 15:33:55 srvl011a kernel: [  725.376875]  ffffffff81064cc7 ffff880fecbeac00 ffff880febfc0100 ffff880febfc0100
Jun 13 15:33:55 srvl011a kernel: [  725.376878]  ffff880fecbeac00 0000000000000000 ffffffff81064db5 ffffffffa0740710
Jun 13 15:33:55 srvl011a kernel: [  725.376881] Call Trace:
Jun 13 15:33:55 srvl011a kernel: [  725.376889]  [<ffffffff814eb7c7>] ? dump_stack+0x41/0x51
Jun 13 15:33:55 srvl011a kernel: [  725.376894]  [<ffffffff81064cc7>] ? warn_slowpath_common+0x87/0xc0
Jun 13 15:33:55 srvl011a kernel: [  725.376896]  [<ffffffff81064db5>] ? warn_slowpath_fmt+0x45/0x50
Jun 13 15:33:55 srvl011a kernel: [  725.376902]  [<ffffffffa073a652>] ? nfsd4_umh_cltrack_init+0x42/0x50 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376907]  [<ffffffffa073b5b1>] ? nfsd4_client_tracking_init+0x91/0x130 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376914]  [<ffffffffa073590d>] ? nfs4_state_start_net+0x27d/0x320 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376919]  [<ffffffffa070fae3>] ? nfsd_svc+0x1b3/0x360 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376924]  [<ffffffffa07108b0>] ? write_filehandle+0x230/0x230 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376927]  [<ffffffff812a008b>] ? simple_strtoull+0x2b/0x50
Jun 13 15:33:55 srvl011a kernel: [  725.376932]  [<ffffffffa07108b0>] ? write_filehandle+0x230/0x230 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376936]  [<ffffffffa0710925>] ? write_threads+0x75/0xc0 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376941]  [<ffffffff81131fa9>] ? __get_free_pages+0x9/0x50
Jun 13 15:33:55 srvl011a kernel: [  725.376946]  [<ffffffff811bd061>] ? simple_transaction_get+0xc1/0xe0
Jun 13 15:33:55 srvl011a kernel: [  725.376951]  [<ffffffffa07108b0>] ? write_filehandle+0x230/0x230 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376955]  [<ffffffffa0710116>] ? nfsctl_transaction_write+0x56/0x80 [nfsd]
Jun 13 15:33:55 srvl011a kernel: [  725.376959]  [<ffffffff81198026>] ? vfs_write+0xc6/0x1f0
Jun 13 15:33:55 srvl011a kernel: [  725.376962]  [<ffffffff8119851b>] ? SyS_write+0x4b/0xb0
Jun 13 15:33:55 srvl011a kernel: [  725.376965]  [<ffffffff814f9779>] ? system_call_fastpath+0x16/0x1b
Jun 13 15:33:55 srvl011a kernel: [  725.376967] ---[ end trace a408a046f76bfc25 ]---
Jun 13 15:33:55 srvl011a kernel: [  725.377001] ------------[ cut here ]------------

I have found a thread on lkml about this (http://lkml.org/lkml/2013/3/1/78),
but it seems there was no conclusion.

Is this something that could be supported?

Yeah, we should really get that fixed. It can probably be worked around by reverting to the old reboot recovery method.

Quoted 4 lines · J. Bruce Fields

Yeah, we should really get that fixed. It can probably be worked around by reverting to the old reboot recovery method.

I would be highly interested. Is the old method still in the kernel sources, maybe supported by a config option?