0000593: Problem with a job on a passive client and limited Maximum Concurrent Jobs

ID	Project	Category	View Status	Date Submitted	Last Update

0000593	bareos-core	General	public	2016-01-07 18:27	2019-12-18 15:25

Reporter	static_void	Assigned To	mvwieringen
Priority	low	Severity	minor	Reproducibility	have not tried
Status	closed	Resolution	fixed
Platform	Linux	OS	Debian	OS Version	7
Product Version	15.2.2
Fixed in Version	15.2.3

Summary	0000593: Problem with a job on a passive client and limited Maximum Concurrent Jobs
Description	Apparently, a job on a passive client takes 2 File Daemon job slots, which I think is not quite reflected in the docs. And Storage Daemon eventually crashes with signal 11 (reporting fatal errors in log) when trying to connect to a passive client when all File Daemon concurrent job slots are taken. My suggestions are: a) reflect that point in the docs, for example in chapter 26.1 "Passive Clients" b) Storage Daemon reaction is too drastic, no need to call this situation a fatal error
Steps To Reproduce	1. Configure a client: Passive = yes (on Director) and Maximum Concurrent Jobs = 1 (on File Daemon) 2. Run a backup job on that client - observe two network connections being established with the File Daemon, but one of those connections have packets queued (shown in netstat, rx-queue on client) - observe job hanging for several minutes - after several minutes, Storage Daemon will report fatal errors concerning authentication, and exit with signal 11 Increasing Maximum Concurrent Jobs on client solves the problem.
Tags	No tags attached.

mvwieringen 2016-01-11 21:08 developer ~0002091	We should probably introduce a variable named MaximumConnections just as what the Director has gotten since the socket server refactoring. This variable then controls the listen backlog and is sized to the right value based on the setting of the number of concurrent jobs allowed. For every fatal error M_FATAL is used this should only abort the Job. So I would be interested to see under a debugger with debug symbols what trap the stored runs into (probably some undetermined code path.)

mvwieringen 2016-01-13 11:13 developer ~0002096	I've added the Maximum Connections parameter to both the SD and FD which is also auto sized when you don't configure it based on the setting of Maximum Concurrent Jobs. We put a set of autogenerated binaries under: http://download.bareos.org/bareos/people/mvw/bug593/ For Debian7 for now as that is what you said you are using. I would still be interested to know where the SD crashes however.

static_void 2016-01-13 11:28 reporter ~0002097	Thanks, going to try the new binaries some time later, maybe today. And I am also gonna try debugging with symbols.

bareos: master b3437eee 2016-01-11 22:37 Marco van Wieringen Ported: N/A Details Diff	socket: Introduce MaximumConnections for filed and stored. When refactoring the socket_server code for the director we added a new config option named MaximumConnections which is used to size the listen backlog. This patch introduces something similar for the filed and stored so we always size the backlog in a proper way according to the wanted concurrency. We now check in accepting a connection in the filed or stored if we already reached the wanted concurrency and then just disconnect the session. Fixes 0000593: Problem with a job on a passive client and limited Maximum Concurrent Jobs	Affected Issues 0000593
	mod - src/filed/dir_cmd.c	Diff File
	mod - src/filed/filed.c	Diff File
	mod - src/filed/filed_conf.c	Diff File
	mod - src/filed/filed_conf.h	Diff File
	mod - src/filed/socket_server.c	Diff File
	mod - src/stored/dir_cmd.c	Diff File
	mod - src/stored/ndmp_tape.c	Diff File
	mod - src/stored/socket_server.c	Diff File
	mod - src/stored/stored.c	Diff File
	mod - src/stored/stored_conf.c	Diff File
	mod - src/stored/stored_conf.h	Diff File

Date Modified	Username	Field	Change
2016-01-07 18:27	static_void	New Issue
2016-01-11 14:49	mvwieringen	Assigned To	=> mvwieringen
2016-01-11 14:49	mvwieringen	Status	new => acknowledged
2016-01-11 21:08	mvwieringen	Note Added: 0002091
2016-01-11 21:08	mvwieringen	Status	acknowledged => feedback
2016-01-13 11:13	mvwieringen	Note Added: 0002096
2016-01-13 11:28	static_void	Note Added: 0002097
2016-01-13 11:28	static_void	Status	feedback => assigned
2016-01-15 16:45	mvwieringen	Changeset attached	=> bareos master b3437eee
2016-01-15 16:46	mvwieringen	Status	assigned => resolved
2016-01-15 16:46	mvwieringen	Fixed in Version	=> 15.2.3
2016-01-15 16:46	mvwieringen	Resolution	open => fixed
2016-02-25 17:01	maik	Relationship added	child of 0000625
2019-12-18 15:25	arogge	Status	resolved => closed