When monitoring /var/log/journal/remote events are missed when two remote hosts send messages at the same time (updates)

Aperta
#134 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
35/100
Tipo di issue
Bug
Chiarezza
Abbastanza chiara
Stato di attività
Ferma
Stack tecnologico
linux, python
Ambito
observability

Direzione di ricerca

Start with the journal.Reader polling example, especially seek_tail(), get_previous(), get_events(), iteration, and process(). Reproduce simultaneous writes from two clients and compare the event counter with journalctl -D /var/log/journal/remote/; done means explaining or correcting the missed-event behavior and confirming that all records are observed.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Context

  • 1 x log server (VM on Apple M2)
  • 2 x log client (VM on Apple M2)
  • Using systemd-journal-remote on server (http)
  • Using systemd-journal-upload on clients (http)

Testing Method
We first cleanup the server with:

rm /var/log/journal/remote/*

On each client we run (side by side / simultaneously:

for x in {1..10000}; do echo "EVENT $x" | logger ; done

So we expect 20,000 records on the server, which we validate:

root@log-server:/home/ansible# journalctl -D /var/log/journal/remote/ --no-pager | grep -v Boot | wc -l
20000

I've tested this many times to confirm that the systemd-journal-remote/upload mechanism is OK.

The failing code

import select
from systemd import journal

 j = journal.Reader(path="/var/log/journal/remote")
 j.seek_tail()
 j.get_previous() # prevent wrap around

 p = select.poll()
 p.register(j, j.get_events())

 print("Starting....")
 event_counter = 0
 while p.poll():
      for x in j:
          event_counter += 1
      if j.process() != journal.APPEND:
          print("...WAITING....")
      else:
          print(f"Event counter: {event_counter}")

First of all, I'm not sure if this code is right. If I've made a glaring mistake, I apologise in advance for this report.

That said, I get correct results if I run my '10K log generator command' on one client at a time.

Event counter: 9915
Event counter: 9933
Event counter: 9945
Event counter: 9950
Event counter: 9987
Event counter: 10000

If I run the command at the same time on each client at once, I would expect 20K records, but it gets nowhere near that:

Event counter: 10228
Event counter: 10228
Event counter: 10228
Event counter: 10228
Event counter: 10228
Event counter: 10228
Event counter: 10228

In this example I'm missing almost half of all expected events. Yet I can confirm on the log server with the 'journalctl' command that there are actually 20K log records stored on the server.

Even if I test with just 1K events per client and run that simultaneously it's also clearly visible:

Event counter: 1096
Event counter: 1141
Event counter: 1215
Event counter: 1430

Happy to run any additional tests if that helps.
If I made a huge mistake, please let me know.

Update 1
I've been testing this issue on virtual machines on my mac.
I've also a few physical x86 micro pc's and the issue is the same.
Rate-limiting the logging with sleep .01 doesn't impact the result.

Question, is python-systemd built to handle remote logging from multiple clients through polling?

Lingua principale
C
Stelle
522
Fork
79
Metriche di merge delle PR
Nessuna PR unita negli ultimi 30g

Guida per i contributori

Nessuna guida per i contributori indicizzata per questo repository

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di systemd/python-systemd

Tutte le issue di systemd/python-systemd

Issue simili

Altre issue su C

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.