Hi Michael
I know/see a lot of activity here, so I assume you are always quite busy and may have missed my last reaction?
Should I proceed with my proposition to add Summary: SUMMARY Services: SERVICES to the pak metadata including an INSTALL_SERVICE_INITSCRIPTS macro
Should I also add InitScripts: INITSCRIPTS for non-service initscripts.. just for completion of metadata ? And in that case a more generic INSTALL_INITSCRIPTS macro accepting a list of initscripts to install? (see mail history below for explanation of my intentions)
Or should we find other methods to have this info ready for services.cgi and/or monitoring agents ?
Regards Robin
Robin Roevens schreef op ma 24-05-2021 om 22:23 [+0200]:
Hi Michael
Michael Tremer schreef op di 18-05-2021 om 16:09 [+0100]:
On 14 May 2021, at 21:07, Robin Roevens robin.roevens@disroot.org wrote:
Hi
Michael Tremer schreef op vr 14-05-2021 om 13:24 [+0100]:
Hello Jonatan,
Yes I get your point. Just because a package has more than one initscript does not mean they both should be listed in services.cgi.
It is a weird corner case, but how do we want to handle this?
-Michael
Maybe a slight renaming of my proposed new variable INITSCRIPTS to SERVICES in the LFS files would make it a bit more clear that this concerns only service-initscripts?
The INSTALL_INITSCRIPTS macro could then be renamed to INSTALL_SERVICE_INITSCRIPTS.
If we then also keep the original INSTALL_INITSCRIPT macro, that macro can then be used to install non-service initscripts if required. Or even to install all required initscripts in the occasion that you want to skip some service-initscripts defined in SERVICES because they are included in the source, like the situation Jonatan pointed out (in this case the SERVICES variable is filled with the sole purpose to serve as metadata).
So this would then result in:
define INSTALL_INITSCRIPT install -m 754 -v $(DIR_SRC)/src/initscripts/packages/$(1) /etc/rc.d/init.d/$(1) endef
define INSTALL_SERVICE_INITSCRIPTS for service in $(SERVICES); do \ $(call INSTALL_INITSCRIPT,$$service) || exit 1; \ done endef
Optionally we could also add again a variable INITSCRIPTS with the purpose to list non-service initscripts. so SERVICES - for service-initscripts INITSCRIPTS - for non-service-initscripts
and in the meta-* files:
Name: NAME Summary: SUMMARY ProgVersion: VER Release: RELEASE Size: SIZE Dependencies: DEPS File: NAME-VER-RELEASE.ipfire Services: SERVICES InitScripts: INITSCRIPTS
Maybe that info could also come in handy sooner or later ?
This sounds rather complicated to me.
What are our objectives here? Does Zabbix want/need to start/stop services? And if so, why?
I was not directly thinking of Zabbix specifically here. The objective is to have a central 'database' of metadata for the pak's that can come in handy for current and possibly future functionality of IPFire. And a single way of retrieving this information (both by using "pakfire info" or directly accessing the pakfire library, in my proposal).
Knowing which pak's provide which services can be used in service.cgi (which currently does give users the possibility to start/stop services indeed) and can be used for monitoring agents like Zabbix, but also Nagios or possibly others to know which services to monitor.
This instead of my earlier proposal of re-using parts of the services.cgi code in a separate script for zabbix_agentd, which you pointed out that should be done cleaner using some central way of storing/retrieving that information.
Zabbix does have functionality to implement starting/stopping those services using a context menu in the Zabbix web GUI; and a user can easily implement that on it's own (as this should be done mainly on server-side anyway) if he wants to and has the information of which services should run. This could be done by zabbix server logging into the IPFire machine using ssh and using addonctrl or by the agent, which would then need permission to addonctrl start/stop. But I think that is something for the user to decide on his own as you pointed out previously, giving zabbix agent root permission on addonctrl start/stop is a possible extra security risk. So I don't think we should provide that out of the box. (Maybe it could be a configurable 'setting' in the IPFire webgui some time in the future, but not something to focus on currently)
But with that said, I don't really understand your remark here as I only propose a way for the LFS files to populate the "Services" meta- data and I only threw in "Initscripts" for non-service initscripts, just for completeness of the meta-data. And to prevent duplication of the list of initscripts, I added a macro to install those in one go instead of again 'defining' this list by needing to call INSTALL_INITSCRIPT individually for each initscript while we now have a list of them in a variable. And due to the remark of Jonatan I now also kept the original INSTALL_INITSCRIPT for the pak maintainer to use in corner cases.
But this has nothing to do with actually stopping/starting services other that providing information about which services are installed (and maybe could be stopped/started if wanted).
If it would make things more clear, I could work my last mail out in an actual patch ?
Regards Robin
In this case the INSTALL_SERVICE_INITSCRIPTS could again be renamed back to INSTALL_INITSCRIPTS, now accepting a parameter which is a list of initscripts to install, so one could call:
$(call INSTALL_INITSCRIPTS,$(SERVICES)) $(call INSTALL_INITSCRIPTS,$(INITSCRIPTS))
to install all initscripts.
Or even provide again an additional macro INSTALL_ALL_INITSCRIPTS which would then install both SERVICES and INITSCRIPTS (or even replace the above suggested INSTALL_SERVICE_INITSCRIPTS, so resulting in a macro to install all initscripts (SERVICES and INITSCRIPTS) and a macro to install a single initscript in case some of the ones listed in SERVICES and/or INITSCRIPTS need to be skipped..
But I think the INSTALL_INITSCRIPTS which takes a list of services to install as parameter is probably the most generic one, giving you the possibility to install all SERVICES in one go, and then individually install only those INITSCRIPTS not included in the source if INITSCRIPTS contains such initscript(s) or vice versa.
Robin
On 12 May 2021, at 19:49, Jonatan Schlag < jonatan.schlag@ipfire.org> wrote:
Hi,
Am 23.04.2021 um 18:16 schrieb Robin Roevens < robin.roevens@disroot.org>:
* Introduce SUMMARY and INITSCRIPTS macro's in LFS- makefiles.
- Add a Summary and InitScripts field to the meta-* addon
files containing the value's of those macro's.
- Replace the INSTALL_INITSCRIPT makefile macro by a new
INSTALL_INITSCRIPTS macro/method that will install all initscripts defined in the new INITSCRIPTS macro.
- Adapt libvirt and zabbix_agentd pak as examples of how to
use this.
Signed-off-by: Robin Roevens robin.roevens@disroot.org
lfs/Config | 8 ++++++-- lfs/libvirt | 6 ++++-- lfs/zabbix_agentd | 5 ++++- src/pakfire/meta | 2 ++ 4 files changed, 16 insertions(+), 5 deletions(-)
diff --git a/lfs/Config b/lfs/Config index eadbbc408..61d6f0c82 100644 --- a/lfs/Config +++ b/lfs/Config @@ -299,15 +299,19 @@ define PAK # Create meta file sed \ -e "s/NAME/$(PROG)/g" \ + -e "s/SUMMARY/$(SUMMARY)/g" \ -e "s/VER/$(VER)/g" \ -e "s/RELEASE/$(PAK_VER)/g" \ -e "s/DEPS/$(DEPS)/g" \ -e "s/SIZE/$$(stat --format=%s /install/packages/$(PROG)- $(VER)-$(PAK_VER).ipfire)/g" \ + -e "s/INITSCRIPTS/$(INITSCRIPTS)/g" \ < /usr/src/src/pakfire/meta > /install/packages/meta- $(PROG) endef
-define INSTALL_INITSCRIPT - install -m 754 -v $(DIR_SRC)/src/initscripts/packages/$(1) /etc/rc.d/init.d/$(1) +define INSTALL_INITSCRIPTS + for initscript in $(INITSCRIPTS); do \ + install -m 754 -v $(DIR_SRC)/src/initscripts/packages/$$initscript /etc/rc.d/init.d/$$initscript; \ + done endef
ifeq "$(BUILD_ARCH)" "$(filter $(BUILD_ARCH),aarch64 riscv64)" diff --git a/lfs/libvirt b/lfs/libvirt index 28a95d317..be5d3db3a 100644 --- a/lfs/libvirt +++ b/lfs/libvirt @@ -25,6 +25,7 @@ include Config
VER = 6.5.0 +SUMMARY = Server side daemon and supporting files for libvirt
THISAPP = libvirt-$(VER) DL_FILE = $(THISAPP).tar.xz @@ -37,6 +38,8 @@ PAK_VER = 25
DEPS = ebtables libpciaccess libtirpc libyajl ncat qemu
+INITSCRIPTS= libvirtd virtlogd
############################################################# ###### ############ # Top-level Rules ############################################################# ###### ############ @@ -121,8 +124,7 @@ $(TARGET) : $(patsubst %,$(DIR_DL)/%,$(objects)) cd $(DIR_APP)/build_libvirt && make install
#install initscripts - $(call INSTALL_INITSCRIPT,libvirtd) - $(call INSTALL_INITSCRIPT,virtlogd) + @$(INSTALL_INITSCRIPTS) mv /usr/libexec/libvirt-guests.sh /etc/rc.d/init.d/libvirt- guests
And here my approach maybe breaks :-). How could we handle this? It is not a daemon, more something like a startup/shutdown scripts for machines... (And should not appear in the service.cgi). So parsing the root file may yield wrong results and your way would be the better one.
Jonatan
# Backup diff --git a/lfs/zabbix_agentd b/lfs/zabbix_agentd index c69643a54..a72fe024b 100644 --- a/lfs/zabbix_agentd +++ b/lfs/zabbix_agentd @@ -25,6 +25,7 @@ include Config
VER = 4.2.6 +SUMMARY = Zabbix Agent
THISAPP = zabbix-$(VER) DL_FILE = $(THISAPP).tar.gz @@ -35,6 +36,8 @@ PROG = zabbix_agentd PAK_VER = 4 DEPS =
+INITSCRIPTS= zabbix_agentd
############################################################# ###### ############ # Top-level Rules ############################################################# ###### ############ @@ -106,7 +109,7 @@ $(TARGET) : $(patsubst %,$(DIR_DL)/%,$(objects)) chown zabbix.zabbix /var/run/zabbix
# Install initscripts - $(call INSTALL_INITSCRIPT,zabbix_agentd) + @$(INSTALL_INITSCRIPTS)
# Install sudoers include file install -v -m 644 $(DIR_SRC)/config/zabbix_agentd/sudoers \ diff --git a/src/pakfire/meta b/src/pakfire/meta index d97b2a0fa..849b9cd6c 100644 --- a/src/pakfire/meta +++ b/src/pakfire/meta @@ -1,6 +1,8 @@ Name: NAME +Summary: SUMMARY ProgVersion: VER Release: RELEASE Size: SIZE Dependencies: DEPS File: NAME-VER-RELEASE.ipfire
+InitScripts: INITSCRIPTS
2.31.1
-- Dit bericht is gescanned op virussen en andere gevaarlijke inhoud door MailScanner en lijkt schoon te zijn.
-- Dit bericht is gescanned op virussen en andere gevaarlijke inhoud door MailScanner en lijkt schoon te zijn.