[Gluster-users] In a replica 2 server, file-updates on one server missing on the other server #Personal#
A Ghoshal
a.ghoshal at tcs.com
Thu Jan 22 08:37:44 UTC 2015
Hi Pranith,
Yes, the very same (chalcogen_eg_oxygen at yahoo.com). Justin Clift sent me a
mail a while back telling me that it is better if we all use our business
email addresses so I made me a new profile.
Glusterfs complains about /proc/sys/net/ipv4/ip_local_reserved_ports
because we use a really old Linux kernel (2.6.34) wherein this feature is
not present. We plan to upgrade our Linux so often but each time we are
dissuaded from it by some compatibility issue or the other. So, we get
this log every time - on both good volumes and bad ones. What bothered me
was this (on serv1):
[2015-01-20 09:37:49.151744] T
[rpc-clnt.c:1182:rpc_clnt_record_build_header] 0-rpc-clnt: Request fraglen
456, payload: 360, rpc hdr: 96
[2015-01-20 09:37:49.151780] T [rpc-clnt.c:1499:rpc_clnt_submit]
0-rpc-clnt: submitted request (XID: 0x39620x Program: GlusterFS 3.3,
ProgVers: 330, Proc: 27) to rpc-transport (replicated_vol-client-0)
[2015-01-20 09:37:49.151810] T [rpc-clnt.c:1302:rpc_clnt_record]
0-replicated_vol-client-1: Auth Info: pid: 7599, uid: 0, gid: 0, owner:
0000000000000000
[2015-01-20 09:37:49.151824] T
[rpc-clnt.c:1182:rpc_clnt_record_build_header] 0-rpc-clnt: Request fraglen
456, payload: 360, rpc hdr: 96
[2015-01-20 09:37:49.151889] T [rpc-clnt.c:1499:rpc_clnt_submit]
0-rpc-clnt: submitted request (XID: 0x39563x Program: GlusterFS 3.3,
ProgVers: 330, Proc: 27) to rpc-transport (replicated_vol-client-1)
[2015-01-20 09:37:49.152239] T [rpc-clnt.c:669:rpc_clnt_reply_init]
0-replicated_vol-client-1: received rpc message (RPC XID: 0x39563x
Program: GlusterFS 3.3, ProgVers: 330, Proc: 27) from rpc-transport
(replicated_vol-client-1)
[2015-01-20 09:37:49.152484] T [rpc-clnt.c:669:rpc_clnt_reply_init]
0-replicated_vol-client-0: received rpc message (RPC XID: 0x39620x
Program: GlusterFS 3.3, ProgVers: 330, Proc: 27) from rpc-transport
(replicated_vol-client-0)
When I write on the good server (serv1), we see that an RPC request is
sent to both client-0 and client-1. While, when I write on the bad server
(serv0), the RPC request is sent only to client-0, which is why it is no
wonder that the writes are not synced over to serv1. Somehow I could not
make the daemon on serv0 understand that there are two up-children and not
just one.
One additional detail - since we are using a kernel that's too old, we do
not have the (Anand Avati's?) FUse readdirplus patches, either. I've
noticed that the fixes in the readdirplus version of glusterfs aren't
always guaranteed to be present on the non-readdirplus version of the
patches. I'd filed a bug around one such anomaly back, but never got
around to writing a patch for it (sorry!) Here it is:
https://bugzilla.redhat.com/show_bug.cgi?id=1062287
Maybe something on similar lines here?
Thanks,
Anirban
P.s. Please ignore the #Personal# in the subject line - we need to do that
to push mails to the public domain past the email filter safely.
From: Pranith Kumar Karampuri <pkarampu at redhat.com>
To: A Ghoshal <a.ghoshal at tcs.com>, gluster-users at gluster.org
Date: 01/22/2015 12:09 AM
Subject: Re: [Gluster-users] In a replica 2 server, file-updates on
one server missing on the other server
hi,
Responses inline.
PS: You are chalkogen_oxygen?
Pranith
On 01/20/2015 05:34 PM, A Ghoshal wrote:
Hello,
I am using the following replicated volume:
root at serv0:~> gluster v info replicated_vol
Volume Name: replicated_vol
Type: Replicate
Volume ID: 26d111e3-7e4c-479e-9355-91635ab7f1c2
Status: Started
Number of Bricks: 1 x 2 = 2
Transport-type: tcp
Bricks:
Brick1: serv0:/mnt/bricks/replicated_vol/brick
Brick2: serv1:/mnt/bricks/replicated_vol/brick
Options Reconfigured:
diagnostics.client-log-level: INFO
network.ping-timeout: 10
nfs.enable-ino32: on
cluster.self-heal-daemon: on
nfs.disable: off
replicated_vol is mounted at /mnt/replicated_vol on both serv0 and serv1.
If I do the following on serv0:
root at serv0:~>echo "cranberries" > /mnt/replicated_vol/testfile
root at serv0:~>echo "tangerines" >> /mnt/replicated_vol/testfile
And then I check for the state of the replicas in the bricks, then I find
that
root at serv0:~>cat /mnt/bricks/replicated_vol/brick/testfile
cranberries
tangerines
root at serv0:~>
root at serv1:~>cat /mnt/bricks/replicated_vol/brick/testfile
root at serv1:~>
As may be seen, the replica on serv1 is blank, when I write into testfile
from serv0 (even though the file is created on both bricks).
Interestingly, if I write something to the file at serv1, then the two
replicas become identical.
root at serv1:~>echo "artichokes" >> /mnt/replicated_vol/testfile
root at serv1:~>cat /mnt/bricks/replicated_vol/brick/testfile
cranberries
tangerines
artichokes
root at serv1:~>
root at serv0:~>cat /mnt/bricks/replicated_vol/brick/testfile
cranberries
tangerines
artichokes
root at serv0:~>
So, I dabbled into the logs a little bit, after upping the diagnostic
level, and this is what I saw:
When I write on serv0 (bad case):
[2015-01-20 09:21:52.197704] T [fuse-bridge.c:546:fuse_lookup_resume]
0-glusterfs-fuse: 53027: LOOKUP
/testfl(f0a76987-8a42-47a2-b027-a823254b736b)
[2015-01-20 09:21:52.197959] D
[afr-common.c:131:afr_lookup_xattr_req_prepare]
0-replicated_vol-replicate-0: /testfl: failed to get the gfid from dict
[2015-01-20 09:21:52.198006] T [rpc-clnt.c:1302:rpc_clnt_record]
0-replicated_vol-client-0: Auth Info: pid: 28151, uid: 0, gid: 0, owner:
0000000000000000
[2015-01-20 09:21:52.198024] T
[rpc-clnt.c:1182:rpc_clnt_record_build_header] 0-rpc-clnt: Request fraglen
456, payload: 360, rpc hdr: 96
[2015-01-20 09:21:52.198108] T [rpc-clnt.c:1499:rpc_clnt_submit]
0-rpc-clnt: submitted request (XID: 0x78163x Program: GlusterFS 3.3,
ProgVers: 330, Proc: 27) to rpc-transport (replicated_vol-client-0)
[2015-01-20 09:21:52.198565] T [rpc-clnt.c:669:rpc_clnt_reply_init]
0-replicated_vol-client-0: received rpc message (RPC XID: 0x78163x
Program: GlusterFS 3.3, ProgVers: 330, Proc: 27) from rpc-transport
(replicated_vol-client-0)
[2015-01-20 09:21:52.198640] D
[afr-self-heal-common.c:138:afr_sh_print_pending_matrix]
0-replicated_vol-replicate-0: pending_matrix: [ 0 3 ]
[2015-01-20 09:21:52.198669] D
[afr-self-heal-common.c:138:afr_sh_print_pending_matrix]
0-replicated_vol-replicate-0: pending_matrix: [ 0 0 ]
[2015-01-20 09:21:52.198681] D
[afr-self-heal-common.c:887:afr_mark_sources]
0-replicated_vol-replicate-0: Number of sources: 1
[2015-01-20 09:21:52.198694] D
[afr-self-heal-data.c:825:afr_lookup_select_read_child_by_txn_type]
0-replicated_vol-replicate-0: returning read_child: 0
[2015-01-20 09:21:52.198705] D
[afr-common.c:1380:afr_lookup_select_read_child]
0-replicated_vol-replicate-0: Source selected as 0 for /testfl
[2015-01-20 09:21:52.198720] D
[afr-common.c:1117:afr_lookup_build_response_params]
0-replicated_vol-replicate-0: Building lookup response from 0
[2015-01-20 09:21:52.198732] D
[afr-common.c:1732:afr_lookup_perform_self_heal]
0-replicated_vol-replicate-0: Only 1 child up - do not attempt to detect
self heal
When I write on serv1 (good case):
[2015-01-20 09:37:49.151506] T [fuse-bridge.c:546:fuse_lookup_resume]
0-glusterfs-fuse: 31212: LOOKUP
/testfl(f0a76987-8a42-47a2-b027-a823254b736b)
[2015-01-20 09:37:49.151683] D
[afr-common.c:131:afr_lookup_xattr_req_prepare]
0-replicated_vol-replicate-0: /testfl: failed to get the gfid from dict
[2015-01-20 09:37:49.151726] T [rpc-clnt.c:1302:rpc_clnt_record]
0-replicated_vol-client-0: Auth Info: pid: 7599, uid: 0, gid: 0, owner:
0000000000000000
[2015-01-20 09:37:49.151744] T
[rpc-clnt.c:1182:rpc_clnt_record_build_header] 0-rpc-clnt: Request fraglen
456, payload: 360, rpc hdr: 96
[2015-01-20 09:37:49.151780] T [rpc-clnt.c:1499:rpc_clnt_submit]
0-rpc-clnt: submitted request (XID: 0x39620x Program: GlusterFS 3.3,
ProgVers: 330, Proc: 27) to rpc-transport (replicated_vol-client-0)
[2015-01-20 09:37:49.151810] T [rpc-clnt.c:1302:rpc_clnt_record]
0-replicated_vol-client-1: Auth Info: pid: 7599, uid: 0, gid: 0, owner:
0000000000000000
[2015-01-20 09:37:49.151824] T
[rpc-clnt.c:1182:rpc_clnt_record_build_header] 0-rpc-clnt: Request fraglen
456, payload: 360, rpc hdr: 96
[2015-01-20 09:37:49.151889] T [rpc-clnt.c:1499:rpc_clnt_submit]
0-rpc-clnt: submitted request (XID: 0x39563x Program: GlusterFS 3.3,
ProgVers: 330, Proc: 27) to rpc-transport (replicated_vol-client-1)
[2015-01-20 09:37:49.152239] T [rpc-clnt.c:669:rpc_clnt_reply_init]
0-replicated_vol-client-1: received rpc message (RPC XID: 0x39563x
Program: GlusterFS 3.3, ProgVers: 330, Proc: 27) from rpc-transport
(replicated_vol-client-1)
[2015-01-20 09:37:49.152484] T [rpc-clnt.c:669:rpc_clnt_reply_init]
0-replicated_vol-client-0: received rpc message (RPC XID: 0x39620x
Program: GlusterFS 3.3, ProgVers: 330, Proc: 27) from rpc-transport
(replicated_vol-client-0)
[2015-01-20 09:37:49.152582] D
[afr-self-heal-common.c:138:afr_sh_print_pending_matrix]
0-replicated_vol-replicate-0: pending_matrix: [ 0 3 ]
[2015-01-20 09:37:49.152596] D
[afr-self-heal-common.c:138:afr_sh_print_pending_matrix]
0-replicated_vol-replicate-0: pending_matrix: [ 0 0 ]
[2015-01-20 09:37:49.152621] D
[afr-self-heal-common.c:887:afr_mark_sources]
0-replicated_vol-replicate-0: Number of sources: 1
[2015-01-20 09:37:49.152633] D
[afr-self-heal-data.c:825:afr_lookup_select_read_child_by_txn_type]
0-replicated_vol-replicate-0: returning read_child: 0
[2015-01-20 09:37:49.152644] D
[afr-common.c:1380:afr_lookup_select_read_child]
0-replicated_vol-replicate-0: Source selected as 0 for /testfl
[2015-01-20 09:37:49.152657] D
[afr-common.c:1117:afr_lookup_build_response_params]
0-replicated_vol-replicate-0: Building lookup response from 0
We see that when you write on serv1, the RPC request is sent to both
replicated_vol-client-0 and replicated_vol-client-1, while when we write
on serv0, the request is sent only to replicated_vol-client-0, and the
FUse client is unaware of the presence of client-1 in the latter case.
I checked a bit more in the logs. When I turn on my trace, I found many
instances of these logs on serv0 but NOT on serv1:
[2015-01-20 09:21:15.520784] T [fuse-bridge.c:681:fuse_attr_cbk]
0-glusterfs-fuse: 53011: LOOKUP() / => 1
[2015-01-20 09:21:17.683088] T [rpc-clnt.c:422:rpc_clnt_reconnect]
0-replicated_vol-client-1: attempting reconnect
[2015-01-20 09:21:17.683159] D [name.c:155:client_fill_address_family]
0-replicated_vol-client-1: address-family not specified, guessing it to be
inet from (remote-host: serv1)
[2015-01-20 09:21:17.683178] T
[name.c:225:af_inet_client_get_remote_sockaddr] 0-replicated_vol-client-1:
option remote-port missing in volume replicated_vol-client-1. Defaulting
to 24007
[2015-01-20 09:21:17.683191] T [common-utils.c:188:gf_resolve_ip6]
0-resolver: flushing DNS cache
[2015-01-20 09:21:17.683202] T [common-utils.c:195:gf_resolve_ip6]
0-resolver: DNS cache not present, freshly probing hostname: serv1
[2015-01-20 09:21:17.683814] D [common-utils.c:237:gf_resolve_ip6]
0-resolver: returning ip-192.168.24.81 (port-24007) for hostname: serv1
and port: 24007
[2015-01-20 09:21:17.684139] D [common-utils.c:257:gf_resolve_ip6]
0-resolver: next DNS query will return: ip-192.168.24.81 port-24007
[2015-01-20 09:21:17.684164] T [socket.c:731:__socket_nodelay]
0-replicated_vol-client-1: NODELAY enabled for socket 10
[2015-01-20 09:21:17.684177] T [socket.c:790:__socket_keepalive]
0-replicated_vol-client-1: Keep-alive enabled for socket 10, interval 2,
idle: 20
[2015-01-20 09:21:17.684236] W [common-utils.c:2247:gf_get_reserved_ports]
0-glusterfs: could not open the file
/proc/sys/net/ipv4/ip_local_reserved_ports for getting reserved ports info
(No such file or directory)
[2015-01-20 09:21:17.684253] W
[common-utils.c:2280:gf_process_reserved_ports] 0-glusterfs: Not able to
get reserved ports, hence there is a possibility that glusterfs may
consume reserved port
Logs above suggest that mount process couldn't assign a reserved port
because it couldn't find the file
/proc/sys/net/ipv4/ip_local_reserved_ports
I guess reboot of the machine fixed it. Wonder why it was not found in the
first place.
Pranith.
[2015-01-20 09:21:17.684660] D [socket.c:605:__socket_shutdown]
0-replicated_vol-client-1: shutdown() returned -1. Transport endpoint is
not connected
[2015-01-20 09:21:17.684699] T
[rpc-clnt.c:519:rpc_clnt_connection_cleanup] 0-replicated_vol-client-1:
cleaning up state in transport object 0x68a630
[2015-01-20 09:21:17.684731] D [socket.c:486:__socket_rwv]
0-replicated_vol-client-1: EOF on socket
[2015-01-20 09:21:17.684750] W [socket.c:514:__socket_rwv]
0-replicated_vol-client-1: readv failed (No data available)
[2015-01-20 09:21:17.684766] D
[socket.c:1962:__socket_proto_state_machine] 0-replicated_vol-client-1:
reading from socket failed. Error (No data available), peer
(192.168.24.81:49198)
I could not find a 'remote-port' option in /var/lib/glusterd on either
peer. Could somebody tell me where this configuration is looked up from?
Also, sometime later, I rebooted serv0 and that seemed to solve the
problem. However, stop+start of replicated_vol and restart of
/etc/init.d/glusterd did NOT solve the problem.
Ignore that log. If no port is given in that volfile, it picks 24007 as
the port, which is the default port where glusterd 'listens'
Any help on this matter will be greatly appreciated as I need to provide
robustness assurances for our setup.
Thanks a lot,
Anirban
P.s. Additional details:
glusterfs version: 3.4.2
Linux kernel version: 2.6.34
=====-----=====-----=====
Notice: The information contained in this e-mail
message and/or attachments to it may contain
confidential or privileged information. If you are
not the intended recipient, any dissemination, use,
review, distribution, printing or copying of the
information contained in this e-mail message
and/or attachments to it are strictly prohibited. If
you have received this communication in error,
please notify us by reply e-mail or telephone and
immediately and permanently delete the message
and any attachments. Thank you
_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://www.gluster.org/mailman/listinfo/gluster-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150122/d34bd906/attachment.html>
More information about the Gluster-users
mailing list