Re: Regression in 3.5.1 or a broken compilation for Debian stretch

2018-02-15 Thread Jose M Calhariz
On Thu, Feb 08, 2018 at 06:06:47PM +, Jean-Louis Martineau wrote:
> add:
> debug-auth 9
> to amanda.conf and retry
> Post the planner debug file.
> 
> For the failing client, do amandad is executed, post the amandad debug files.
> 
> Jean-Louis

As promised I an posting here a summary of the private emails exchanged
between me and Jean-Louis.

After posting my debug logs Jean-Louis have quickly identified a quick
fix.  The patch is attached, in case other people may find it usefull
before a new release of amanda.

Kind regards
Jose M Calhariz


-- 
--

Não tenha pressa, mas não perca tempo

--José Saramago
diff --git a/common-src/protocol.c b/common-src/protocol.c
index 1ab7294..972ef4e 100644
--- a/common-src/protocol.c
+++ b/common-src/protocol.c
@@ -179,8 +179,8 @@ protocol_sendreq(
 void *			datap)
 {
 proto_t *p;
-char*platform = NULL;
-char*distro = NULL;
+static char *platform = NULL;
+static char *distro = NULL;
 
 p = g_malloc(sizeof(proto_t));
 p->state = s_sendreq;
@@ -209,7 +209,10 @@ protocol_sendreq(
 proto_debug(1, _("protocol: security_connect: host %s -> p %p\n"),
 		hostname, p);
 
-get_platform_and_distro(&platform, &distro);
+if (!platform && !distro) {
+	get_platform_and_distro(&platform, &distro);
+}
+
 if (distro != NULL &&
 	!g_str_equal(distro, "mac") &&
 #if defined HAVE_FUNC_GETSERVBYNAME_R_4 || defined HAVE_FUNC_GETSERVBYNAME_R_5 || defined HAVE_FUNC_GETSERVBYNAME_R_6
@@ -243,8 +246,6 @@ protocol_sendreq(
 	security_connect(p->security_driver, p->hostname, p->conf_fn, connect_callbackX,
 			 p, p->datap);
 }
-g_free(platform);
-g_free(distro);
 }
 
 static gpointer


Re: Regression in 3.5.1 or a broken compilation for Debian stretch

2018-02-08 Thread Jean-Louis Martineau
add:
debug-auth 9
to amanda.conf and retry
Post the planner debug file.

For the failing client, do amandad is executed, post the amandad debug files.

Jean-Louis


From: owner-amanda-us...@amanda.org  on behalf 
of Jose M Calhariz 
Sent: February 8, 2018 10:30:11 AM
To: amanda-users@amanda.org
Subject: Regression in 3.5.1 or a broken compilation for Debian stretch


Hi,

I have been testing the new amanda 3.5 and 3.5.1 on Debian, on my
servers, before uploading the package into to Debian sid.

The 3.5 successfully solved the problem with big installations of
amanda here the fault of one or 2 clients cascade to more failures
during backups or "amcheck -c".

I have been testing the 3.5.1 in my little server with success and
upload it to Debian. But now that I have installed it on my big
server, 130 clients and 700 DLE, I am seeing it again the cascade
problem of multiple failures because of one or two failures. I am
seeking help to pinpoint the problem. It can be a regression on the
code or must probably a problem on my compiled packages for stretch.

Main question: What logs I should look or what patch I can apply to
try to identified the root cause of the problem?


Details

Currently I have 3 hosts that are unreachable. This cause a non
predictable failure. Meaning that sometimes everything works as
expected, but most of the times more clients fail with:

selfcheck request failed: error sending REQ: write error to: Broken
pipe

Looking into one of the clients that fail, I do not see an obvious
error message in the amanda logs.

I am willing to continue looking into the logs or running modified
code or even run a debugger. For privacy reasons I will post here a
resume of the findings or send by private email the full logs and
other results.


More details:

The server runs Debian 9 (stretch) and use amanda software 3.5.1 from
Debian sid recompiled for stretch. Most of the clients are Debian 9
and runs amanda client from Debian 3.3.9-5 or 3.5 and 3.5.1
(backported from Debian sid).

The patch that I have applied for Debian should not have changed this
behaviour and can be checked in:

https://packages.debian.org/buster/amanda-server<https://packages.debian.org/buster/amanda-server>

TIA
Kind regards
Jose M Calhariz


--
--
Goze a vida, afinal você nasceu de uma gozada!
This message is the property of CARBONITE, INC. and may contain confidential or 
privileged information.
If this message has been delivered to you by mistake, then do not copy or 
deliver this message to anyone.  Instead, destroy it and notify me by reply 
e-mail


Regression in 3.5.1 or a broken compilation for Debian stretch

2018-02-08 Thread Jose M Calhariz

Hi,

I have been testing the new amanda 3.5 and 3.5.1 on Debian, on my
servers, before uploading the package into to Debian sid.

The 3.5 successfully solved the problem with big installations of
amanda here the fault of one or 2 clients cascade to more failures
during backups or "amcheck -c".

I have been testing the 3.5.1 in my little server with success and
upload it to Debian.  But now that I have installed it on my big
server, 130 clients and 700 DLE, I am seeing it again the cascade
problem of multiple failures because of one or two failures.  I am
seeking help to pinpoint the problem.  It can be a regression on the
code or must probably a problem on my compiled packages for stretch.

Main question:  What logs I should look or what patch I can apply to
try to identified the root cause of the problem?


 Details

Currently I have 3 hosts that are unreachable.  This cause a non
predictable failure.  Meaning that sometimes everything works as
expected, but most of the times more clients fail with: 

selfcheck request failed: error sending REQ: write error to: Broken
pipe

Looking into one of the clients that fail, I do not see an obvious
error message in the amanda logs.

I am willing to continue looking into the logs or running modified
code or even run a debugger.  For privacy reasons I will post here a
resume of the findings or send by private email the full logs and
other results.


 More details:

The server runs Debian 9 (stretch) and use amanda software 3.5.1 from
Debian sid recompiled for stretch.  Most of the clients are Debian 9
and runs amanda client from Debian 3.3.9-5 or 3.5 and 3.5.1
(backported from Debian sid).

The patch that I have applied for Debian should not have changed this
behaviour and can be checked in:

https://packages.debian.org/buster/amanda-server 

TIA
Kind regards
Jose M Calhariz


-- 
--
Goze a vida, afinal você nasceu de uma gozada!