glusterfs.git/rpc/rpc-transport/socket/src, branch v3.8rc2

rpc: change client insecure port ceiling from 65535 to 49151

2016-05-19T08:56:53+00:00

current port allocation to various processes (clumsy):

 1023 - 1       -> client ports range if bind secure is turned on
49151 - 1024    -> fall back to this, if in above case ports exhaust
65535 - 1024    -> client port range if bind insecure is on
49152 - 65535   -> brick port range

now, we have segregated port ranges 0 - 65535 to below 3 ranges

 1023 - 1       -> client ports range if bind secure is turned on
49151 - 1024    -> client port range if bind insecure is on
                   (fall back to this, if in above case ports exhaust)
49152 - 65535   -> brick port range

so now we have a clean segregation of port mapping

Backport of:
> Change-Id: Ie3b4e7703e0bbeabbe0adbdd6c60d9ef78ef7c65
> BUG: 1335776
> Signed-off-by: Prasanna Kumar Kalever 
> Reviewed-on: http://review.gluster.org/14326
> Tested-by: Prasanna Kumar Kalever 
> Reviewed-by: Raghavendra Talur 
> Tested-by: Gluster Build System 
> CentOS-regression: Gluster Build System 
> NetBSD-regression: NetBSD Build System 
> Smoke: Gluster Build System 
> Reviewed-by: Kaleb KEITHLEY 
> Reviewed-by: Raghavendra G 

Change-Id: Ie3b4e7703e0bbeabbe0adbdd6c60d9ef78ef7c65
BUG: 1337127
Signed-off-by: Prasanna Kumar Kalever 
Reviewed-on: http://review.gluster.org/14413
Tested-by: Prasanna Kumar Kalever 
Smoke: Gluster Build System 
CentOS-regression: Gluster Build System 
NetBSD-regression: NetBSD Build System 
Reviewed-by: Raghavendra G

socket: Fix incorrect handling of partial reads

2016-05-12T05:17:25+00:00

The usage of function local variables in the protocol state
machine caused an incorrect behaviour when a partial read
from the socket forced the function to return and restart
later when more data was available. At this point the local
variables contained incorrect data.

> Change-Id: I4db1f4ef5c46a3d2d7f7c5328e906188c3af49e6
> BUG: 1334285
> Signed-off-by: Xavier Hernandez 
> Reviewed-on: http://review.gluster.org/14270
> Reviewed-by: Raghavendra G 
> Smoke: Gluster Build System 
> NetBSD-regression: NetBSD Build System 
> Tested-by: Raghavendra G 
> CentOS-regression: Gluster Build System 

Change-Id: I92b7c91b4c0dfc15224aea39308c93b27028dd4f
BUG: 1334287
Signed-off-by: Xavier Hernandez 
Reviewed-on: http://review.gluster.org/14293
NetBSD-regression: NetBSD Build System 
Reviewed-by: Raghavendra G 
CentOS-regression: Gluster Build System 
Smoke: Gluster Build System

glusterd: add defence mechanism to avoid brick port clashes

2016-05-10T07:11:33+00:00

Intro:
Currently glusterd maintain the portmap registry which contains ports that
are free to use between 49152 - 65535, this registry is initialized
once, and updated accordingly as an then when glusterd sees they are been
used.

Glusterd first checks for a port within the portmap registry and gets a FREE
port marked in it, then checks if that port is currently free using a connect()
function then passes it to brick process which have to bind on it.

Problem:
We see that there is a time gap between glusterd checking the port with
connect() and brick process actually binding on it. In this time gap it could
be so possible that any process would have occupied this port because of which
brick will fail to bind and exit.

Case 1:
To avoid the gluster client process occupying the port supplied by glusterd :

we have separated the client port map range with brick port map range more @
http://review.gluster.org/#/c/13998/

Case 2: (Handled by this patch)
To avoid the other foreign process occupying the port supplied by glusterd :

To handle above situation this patch implements a mechanism to return EADDRINUSE
error code to glusterd, upon which a new port is allocated and try to restart
the brick process with the newly allocated port.

Note: Incase of glusterd restarts i.e. runner_run_nowait() there is no way to
handle Case 2, becuase runner_run_nowait() will not wait to get the return/exit
code of the executed command (brick process). Hence as of now in such case,
we cannot know with what error the brick has failed to connect.

This patch also fix the runner_end() to perform some cleanup w.r.t
return values.

Backport of:
> Change-Id: Iec52e7f5d87ce938d173f8ef16aa77fd573f2c5e
> BUG: 1322805
> Signed-off-by: Prasanna Kumar Kalever 
> Reviewed-on: http://review.gluster.org/14043
> Tested-by: Prasanna Kumar Kalever 
> Reviewed-by: Atin Mukherjee 
> Smoke: Gluster Build System 
> NetBSD-regression: NetBSD Build System 
> CentOS-regression: Gluster Build System 
> Reviewed-by: Raghavendra G 

Change-Id: Id7d8351a0082b44310177e714edc0571ad0f7195
BUG: 1333711
Signed-off-by: Prasanna Kumar Kalever 
Reviewed-on: http://review.gluster.org/14235
Tested-by: Prasanna Kumar Kalever 
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Atin Mukherjee

rpc: define client port range

2016-05-10T07:11:20+00:00

Problem:
when bind-insecure is 'off', all the clients bind to secure ports,
if incase all the secure ports exhaust the client will no more bind
to secure ports and tries gets a random port which is obviously insecure.

we have seen the client obtaining a port number in the range 49152-65535
which are actually reserved as part of glusterd's pmap_registry for bricks,
hence this will lead to port clashes between client and brick processes.

Solution:
If we can define different port ranges for clients incase where secure ports
exhaust, we can avoid the maximum port clashes with in gluster processes.

Still we are prone to have clashes with other non-gluster processes, but
the chances being very low, but that's a different story on its own,
which will be handled in upcoming patches.

Backportof:
> Change-Id: Ib5ce05991aa1290ccb17f6f04ffd65caf411feaf
> BUG: 1322805
> Signed-off-by: Prasanna Kumar Kalever 
> Reviewed-on: http://review.gluster.org/13998
> Smoke: Gluster Build System 
> NetBSD-regression: NetBSD Build System 
> CentOS-regression: Gluster Build System 
> Reviewed-by: Atin Mukherjee 
> Reviewed-by: Raghavendra G 

Change-Id: I2ab9608ddbefcdf5987d817c23dd066010148e19
BUG: 1333711
Signed-off-by: Prasanna Kumar Kalever 
Reviewed-on: http://review.gluster.org/14234
Tested-by: Prasanna Kumar Kalever 
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Atin Mukherjee

socket: Reap own-threads

2016-04-29T13:34:53+00:00

Dead own-threads are reaped periodically (currently every minute). This
helps avoid memory being leaked, and should help prevent memory
starvation issues with GlusterD.

Change-Id: Ifb3442a91891b164655bb2aa72210b13cee31599
BUG: 1331289
Signed-off-by: Kaushal M 
Reviewed-on: http://review.gluster.org/14101
Smoke: Gluster Build System 
Tested-by: Jeff Darcy 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Jeff Darcy

rpc: fix gf_process_reserved_ports

2016-04-27T15:05:08+00:00

this patch also does minor code cleanups.

Change-Id: I0d005bd0f9baaaae498aa1df4faa6fcb65fa7a6e
BUG: 1198849
Signed-off-by: Prasanna Kumar Kalever 
Reviewed-on: http://review.gluster.org/13997
Tested-by: Prasanna Kumar Kalever 
Smoke: Gluster Build System 
Reviewed-by: Atin Mukherjee 
CentOS-regression: Gluster Build System 
Reviewed-by: Jeff Darcy 
NetBSD-regression: NetBSD Build System

socket: Don't cleanup encrypted transport in socket_connect()

2016-04-08T01:06:25+00:00

..instead cleanup only in socket_poller()

With commit d117466 socket_poller() wasn't launched from socket_connect
(for encrypted connections), if connect() failed. This was done to
prevent the socket private data from being double unreffed, from the
cleanups in both socket_poller() and socket_connect(). This allowed
future reconnects to happen successfully.

If a socket reconnects is sort of decided by the rpc notify function
registered. The above change worked with glusterd, as the glusterd rpc
notify function (glusterd_peer_rpc_notify()) continuously allowed
reconnects on failure.

mgmt_rpc_notify(), the rpc notify function in glusterfsd, behaves
differently.

For a DISCONNECT event, if more volfile servers are available or if more
addresses are available in the dns cache, it allows reconnects. If not
it terminates the program.

For a CONNECT event, it attempts to do a volfile fetch rpc request. If
sending this rpc fails, it immediately terminates the program.

One side effect of commit d117466, was that the encrypted socket was
registered with epoll, unintentionally, on a connect failure.  A weird
thing happens because of this. The epoll notifier notifies
mgmt_rpc_notify() of a CONNECT event, instead of a DISCONNECT as
expected. This causes mgmt_rpc_notify() to attempt an unsuccessful
volfile fetch rpc request, and terminate.
(I still don't know why the epoll raises the CONNECT event)

Commit 46bd29e fixed some issues with IPv6 in GlusterFS. This caused
address resolution in GlusterFS to also request of IPv6 addresses
(AF_UNSPEC) instead of just IPv4. On most systems, this causes the IPv6
addresses to be returned first.

GlusterD listens on 0.0.0.0:24007 by default. While this attaches to all
interfaces, it only listens on IPv4 addresses. GlusterFS daemons and
bricks are given 'localhost' as the volfile server. This resolves to
'::1' as the first address.

When using management encryption, the above reasons cause the daemon
processes to fail to fetch volfiles and terminate.

Solution
--------
The solution to this is simple. Instead of cleaning up the encrypted
socket in socket_connect(), launch socket_poller() and let it cleanup
the socket instead. This prevents the unintentional registration with
epoll, and socket_poller() sends the correct events to the rpc notify
functions, which allows proper reconnects to happen.

Change-Id: Idb0c0a828743cccca51cfdd1aa6458cfa0a9d100
Signed-off-by: Kaushal M 
Reviewed-on: http://review.gluster.org/13926
Smoke: Gluster Build System 
Tested-by: Jeff Darcy 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Jeff Darcy

rpc: assign port only if it is unreserved

2016-04-06T17:08:41+00:00

Current order:
    assign port;
    check for port;
    if reserved {
        port--;
        continue to i;
    }
    bind();

basically, we are assigning port first then checking if it is reserved

Fix:
    get unreserved port;
    assign port;
    bind();

from now, we get unreserved port first and then assign it

Change-Id: I004580c5215e7c9cae3594af6405b20fcd9fa4ad
BUG: 1323659
Signed-off-by: Prasanna Kumar Kalever 
Reviewed-on: http://review.gluster.org/13900
Tested-by: Prasanna Kumar Kalever 
NetBSD-regression: NetBSD Build System 
Reviewed-by: Jeff Darcy 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra Talur 
Smoke: Gluster Build System 
Reviewed-by: Raghavendra G

socket: Launch socket_poller only if connect succeeded

2016-03-04T04:42:14+00:00

For an encrypted connection, sockect_connect() used to launch
socket_poller() in it's own thread (ON by default), even if the connect
failed. This would cause two unrefs to be done on the transport, once in
socket_poller() and once in socket_connect(), causing the transport to
be freed and cleaned up. This would cause further reconnect attempts
from failing as the transport wouldn't be available.

By starting socket_poller() only if connect succeeded, this is avoided.

Change-Id: Ie22090dbb1833bdd0f636a76cb3935d766711917
BUG: 1313206
Signed-off-by: Kaushal M 
Reviewed-on: http://review.gluster.org/13554
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra G

socket: reduce rate of readv failure logs due to disconnect

2016-02-22T18:10:50+00:00

... by using GF_LOG_OCCASIONALLY

Change-Id: I779ff32ead13c8bb446a57b5baccf068ae992df1
BUG: 1114847
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.org/8210
Tested-by: Atin Mukherjee 
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra G