RHCSA V10
Part of RHCSA V10. These aren’t two independent, parallel loggers — systemd-journald is the actual collection point for everything (kernel messages, early boot output, every daemon’s stdout/stderr, and syslog events), and rsyslog sits downstream of it, reading from the journal (via its imjournal module) to sort messages into the traditional text files under /var/log/. journald’s own copy is structured/binary and — by default — not persistent across reboots; rsyslog’s text files in /var/log/ are what actually survives long-term, which is exactly why “make the journal persistent” (below) is worth doing rather than relying on rsyslog’s copy alone.
journald / journalctl
Every unit systemd manages sends its output to the journal — this is the log source on a modern RHEL box, including for services with no separate log file of their own.
journalctl # entire journal, oldest first
journalctl -e # jump to the end (newest entries) — like tail
journalctl -f # follow live, like tail -f
journalctl -r # reverse order, newest firstFiltering — this is where journalctl earns its keep
journalctl -u sshd # only this systemd unit's logsJul 30 08:14:02 host sshd[1842]: Server listening on 0.0.0.0 port 22.
Jul 30 08:41:17 host sshd[2093]: Accepted publickey for kingsley from 10.0.0.5 port 51422 ssh2
Jul 30 09:02:44 host sshd[2211]: Failed password for invalid user admin from 203.0.113.9 port 39102 ssh2
Each line is exactly what it would’ve looked like in /var/log/secure — journalctl -u <unit> is the modern, no-separate-log-file-needed replacement for “grep the right log file for this service’s name.”
journalctl -u sshd -f # follow just this unit live
journalctl -b # only since the current boot
journalctl -b -1 # the PREVIOUS boot — critical for "why did it crash last night"
journalctl --list-boots # see all boots the journal has recorded, with their offsets
journalctl --since "2026-07-23 09:00" --until "2026-07-23 10:00"
journalctl --since yesterday
journalctl -k # kernel messages only (like dmesg, but persistent across reboots if journal is persistent)
journalctl -p err # priority filter — err and everything more severe
journalctl -p warning..err # priority range
journalctl -o json-pretty # structured output — useful for scripting/piping to jq
journalctl -o verbose # every indexed field for each entry, not just the summary line — see belowFiltering by field — more precise than -u when you need it
Every journal entry carries structured fields, not just a text message — journalctl -o verbose reveals them (_PID, _UID, _COMM, _EXE, _SYSTEMD_UNIT, and more; the full list is in systemd.journal-fields(7)). You can filter directly on any of them, and combine several at once (combining fields is an implicit AND):
journalctl _PID=1 # everything logged by PID 1 specifically (systemd itself)
journalctl _UID=81 # everything logged by a specific UID
journalctl _SYSTEMD_UNIT=sshd.service _PID=2188 # narrow to one unit AND one specific process instance of itThis is the tool for “which specific process instance logged this,” in a way -u <unit> alone can’t answer once a unit has restarted several times with different PIDs.
Priority levels, from most to least severe (numeric values you’ll see in -p output or scripting):
| # | Name |
|---|---|
| 0 | emerg |
| 1 | alert |
| 2 | crit |
| 3 | err |
| 4 | warning |
| 5 | notice |
| 6 | info |
| 7 | debug |
Persistent vs. volatile journal storage
By default on a lot of installs, the journal only lives in /run/log/journal — RAM-backed, wiped on every reboot. For a real server you almost always want it to survive reboots:
mkdir -p /var/log/journal
systemd-tmpfiles --create --prefix /var/log/journal
systemctl restart systemd-journaldOr set it explicitly in /etc/systemd/journald.conf:
[Journal]
Storage=persistentThis exact task — “make the journal persistent” — is a classic RHCSA-style exam item because it’s a one-liner if you know it and a total guess if you don’t.
Managing journal size
journalctl --disk-usage # how much space the journal is currently using
journalctl --vacuum-size=500M # shrink journal down to this size, deleting oldest entries
journalctl --vacuum-time=2weeks # delete anything older than 2 weeksrsyslog — the text-file layer
Still installed alongside journald on RHEL and responsible for the traditional /var/log/* text files a lot of tools and admins still expect.
/etc/rsyslog.conf # main config
/etc/rsyslog.d/*.conf # drop-in configs, same pattern as sudoers.d/logrotate.dRules follow a facility.priority destination format:
mail.* /var/log/maillog
*.info;mail.none;cron.none /var/log/messages
authpriv.* /var/log/secure
cron.* /var/log/cron
Common log files this produces on RHEL:
| File | Contents |
|---|---|
/var/log/messages | General system messages — the default catch-all |
/var/log/secure | Authentication and authorization events — sudo, SSH logins, su |
/var/log/cron | Cron job execution log |
/var/log/maillog | Mail subsystem |
/var/log/boot.log | Boot-time service startup messages |
systemctl restart rsyslog # apply config changeslogrotate — keeping logs from eating the disk
Runs daily via a systemd timer (logrotate.timer), rotating/compressing/deleting old log files per rules.
/etc/logrotate.conf # global defaults
/etc/logrotate.d/* # per-service overrides — most packages drop one of these in on installExample drop-in:
/var/log/myapp/*.log {
weekly
rotate 4
compress
missingok
notifempty
size 50M
}
| Directive | Meaning |
|---|---|
weekly / daily / monthly | Rotation frequency |
rotate N | Keep N old rotated copies before deleting the oldest |
compress | Gzip rotated files |
missingok | Don’t error if the log file doesn’t exist |
notifempty | Skip rotation if the log is empty |
size 50M | Rotate once the file crosses this size, regardless of schedule |
logrotate -f /etc/logrotate.conf # force a rotation run right now — useful for testing a rule without waiting for the timer
logrotate -d /etc/logrotate.conf # dry run, shows what WOULD happenA note on timestamps
Logs are only as useful as their timestamps are accurate. Worth a quick sanity check with timedatectl and, if NTP (Network Time Protocol) sync via chronyd is off, logs across a fleet of machines can drift out of correlation with each other — a real problem when you’re trying to line up an error across a web server and a database server, where “which happened first” is the entire question you’re trying to answer.
timedatectl # current time, timezone, whether NTP sync is active
chronyc sources # what time servers chronyd is actually syncing againstSoftware engineering
Infrastructure & security
Computer science
Hardware & engineering