Hi Ted, I will collect some info about jbd2 and get back to this list. To answer your questions, our workload is mostly sequential writes from a couple of daemons. Every second, we will create 6 to 10 files, and write around 8M bytes into them. We only do extend or truncate when needed, but I am sure we rarely do this on NVMe devices. Reading the kernel document, I also see ext4 ordered option will combine metadata and data blocks for journaling. If the data block is big, and the workload is heavy, will JBD2 be soaked? Thanks, Roy ---------------------------------------- > Date: Wed, 12 Aug 2015 14:37:50 -0400 > From: tytso@xxxxxxx > To: bsdnet@xxxxxxxxxxx > CC: linux-ext4@xxxxxxxxxxxxxxx > Subject: Re: fsync stuck at jbd2_log_wait_commit on NVMe devices > > Hi Roy, > > My suggestion is to collect information from the jbd2_run_stats and > jbd2_checkpoint_stats tracepoints. > > cd /sys/kernel/debug/tracing > echo 1> events/jbd2/jbd2_run_stats/enable > echo 1> events/jbd2/jbd2_checkpoint_stats/enable > cat trace_pipe> /tmp/traces & > tail -f /tmp/traces > > The jbd2_handle_stats tracepoint can be informative, but it's also far > more voluminous. > > That will give us a hint where things are getting bottlenecked. > > What sort of workload is your application doing? Is it just primarily > doing random writes into a preallocated file? Is it creating or > deleting files? Extending or truncating files? etc. > > - Ted > -- > To unsubscribe from this list: send the line "unsubscribe linux-ext4" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
