Make sure journald never blocks on sd_notify() to PID 1 #1745
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This adds support for a new environment variable SYSTEMCTL_INSTALL_CLIENT_SIDE, that ensures that systemctl executes install operations client-side instead of passing them to PID1. This is useful in debugging situations, but even beyond that. However, we don't want to make it official API, hence let's just make it an undocumented environment variable. Similar, add a second variable, SYSTEMCTL_SKIP_SYSV which allows skipping the SysV chkconfig fall-back if set. This is useful for similar reasons, and exposed as undocumented as environment variable for similar reasons, too.
When rebooting nspawn containers about 400 times we'd otherwise hit the fd limit and refuse further reboots.
Let's make sure we don't even try to create the audit socket
- Make sure we log each error at least once, and at most once - Replace FOREACH_WORD loops by extract_first_word() loops - Use FOREACH_DIRENT() for directory loops - Use free_and_strdup() where appropriate - Do not operate on half-loaded SysV files - Always properly free all memory
The default of 16 is pretty low, let's bump this to accomodate for more queued datagrams. This is useful for AF_UNIX/SOCK_DGRAM logging and sd_notify() sockets as this allows queuing more datagrams before things start to block, thus improving parallelization and logging performance.
Let's make sure we don't start blocking on sd_notify() earlier than necessary, let's bump the socket buffer sizes to 8M. We already do something similar for our logging socket buffers, hence apply a similar bump here.
Better generate a real error then simply connect to the wrong socket.
Otherwise we might run into deadlocks, when journald blocks on the notify socket on PID 1, and PID 1 blocks on IPC to dbus-daemon and dbus-daemon blocks on logging to journald. Break this cycle by making sure that journald never ever blocks on PID 1. Note that this change disables support for event loop watchdog support, as these messages are sent in blocking style by sd-event. That should not be a big loss though, as people reported frequent problems with the watchdog hitting journald on excessively slow IO. Fixes: systemd#1505.
|
I ran the "boot-smoke" test successfully with 50 reboots, so this fixes this nasty bug \o/. I also ran the complete integration test suite on this (success too). 👍 Thanks! |
|
Yeah, bumping the sysctl is difficult, I figure we should do that from PID 1 much earlier even, as neither the logging socket, nor the notify socket are actually affected by it now. |
zonque
added a commit
that referenced
this pull request
Nov 2, 2015
Make sure journald never blocks on sd_notify() to PID 1
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is a fix for #1505.
This is on top of #1737, hence please review and merge that one first. And ignore the first commits, only the last three are actually really new in this PR and are the ones to review.