On Tue, Nov 19, 2019 at 04:57:46PM +0000, Martin Wilck wrote:
On Tue, 2019-11-19 at 17:32 +0100, Michal Suchánek wrote:
On Tue, Nov 19, 2019 at 03:25:19PM +0000, Martin Wilck wrote:
On Tue, 2019-11-19 at 15:47 +0100, Stefan Seyfried wrote:
I've come to thoroughly hate it since I had to tame landscapes with SOC6 and SOC7 (where the journal was in RAM only, but still the ridiculous amount of logging in openstack combined with the ridiculous performance of journalctl made running "supportconfig" an 8hours+ job. Most of the time was spent in something like "journalctl -b > journal.txt". I believe we even got a PTF for supportutils working around the issue).
Yuck. Looks like a thing someone should fix, some time. Have you considered filing a bug or feature request?
State of the art fix is 'zypper in rsyslog'. Then you can use cp which has performance orders of magnitude better than journalctl.
I don't think fixing systemd journal design is something somebody would seriously consider by now. It is not like nobody has tried, is it?
I haven't followed the development close enough to tell. But it's such a central element of modern Linux distros that it's a real shame it performs so badly.
I would say it is not dissimilar to what we have with Spectre. If you want decent performance you need a CPU that has this issue, and if you care you apply a workaround. If you want modern system with all desktop services you need systemd that has this issue, and if you care you apply workaround. In either case persuading upstream to fix the issue was unsuccessful so far. Unlike with journal there is at least some progress with Spectre. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org