mbox series

[ovs-dev,00/11] Get rid of ctl_fatal() calls in ovn-nbctl (part 2)

Message ID 20180717133415.23781-1-jkbs@redhat.com
Headers show
Series Get rid of ctl_fatal() calls in ovn-nbctl (part 2) | expand

Message

Jakub Sitnicki July 17, 2018, 1:34 p.m. UTC 
This is a continuation of an earlier series that aims to replace calls to
ctl_fatal() in command handlers in ovn-nbctl. The motivation is to handle errors
gracefully when running commands in daemon mode because as a long-lived process
we shouldn't terminate on errors that we can recover from.

After this series there are no more ctl_fatal() calls in ovn-nbctl that affect
the daemon mode. The only remaining function left to convert is the commands
parser in db-ctl-base module (ctl_parse_commands()), which I intend to deal with
separately. Either as a part of ovn-nbctl daemon series (already in review [1]),
or as a follow-up to it.

Thanks,
Jakub

[1] https://patchwork.ozlabs.org/project/openvswitch/list/?series=55472

Jakub Sitnicki (11):
  ovn-nbctl: Don't die in pg_by_name_or_uuid().
  ovn-nbctl: Don't die in gc_by_name_or_uuid().
  ovn-nbctl: Don't die in lsp_to_ls().
  ovn-nbctl: Don't die in lrp_to_lr().
  ovn-nbctl: Don't die in parse_priority().
  ovn-nbctl: Don't die in parse_direction().
  ovn-nbctl: Don't die in dhcp_options_get().
  ovn-nbctl: Propagate error thru the context.
  ovn-nbctl: Use ctl_error() in command handlers.
  ovn-nbctl: Remove pointless "return;" at ends of functions.
  ovn-nbctl: Fix mem leak in nbctl_lrp_set_gateway_chassis().

 ovn/utilities/ovn-nbctl.c | 360 ++++++++++++++++++++++++++++++++--------------
 tests/ovn-nbctl.at        |   5 +
 2 files changed, 260 insertions(+), 105 deletions(-)

--
2.14.4

Comments

Ben Pfaff July 23, 2018, 10:46 p.m. UTC | #1 
On Tue, Jul 17, 2018 at 03:34:04PM +0200, Jakub Sitnicki wrote:
> This is a continuation of an earlier series that aims to replace calls to
> ctl_fatal() in command handlers in ovn-nbctl. The motivation is to handle errors
> gracefully when running commands in daemon mode because as a long-lived process
> we shouldn't terminate on errors that we can recover from.
> 
> After this series there are no more ctl_fatal() calls in ovn-nbctl that affect
> the daemon mode. The only remaining function left to convert is the commands
> parser in db-ctl-base module (ctl_parse_commands()), which I intend to deal with
> separately. Either as a part of ovn-nbctl daemon series (already in review [1]),
> or as a follow-up to it.
> 
> Thanks,
> Jakub
> 
> [1] https://patchwork.ozlabs.org/project/openvswitch/list/?series=55472

I applied this series to master.  Thank you!