Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
From:   Max Kellermann <max.kellermann@ionos.com>
To:     Xiubo Li <xiubli@redhat.com>, Ilya Dryomov <idryomov@gmail.com>,
        Jeff Layton <jlayton@kernel.org>
Cc:     Max Kellermann <max.kellermann@ionos.com>,
        ceph-devel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: [PATCH 2/2] fs/ceph/debugfs: expose raw metric counters
Date:   Fri, 22 Sep 2023 08:25:58 +0200
Message-Id: <20230922062558.1739642-2-max.kellermann@ionos.com>
In-Reply-To: <20230922062558.1739642-1-max.kellermann@ionos.com>
References: <20230922062558.1739642-1-max.kellermann@ionos.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

To enable userspace to calculate the current latency, not just the
average latency since the filesystem was mounted.

We have been running this patch for a while on our servers and our
Prometheus exporter collects these statistics:

 https://github.com/CM4all/Prometheus-Exporters/
 https://github.com/CM4all/Prometheus-Exporters/blob/master/src/KernelExporter.cxx

Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
---
 fs/ceph/debugfs.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
index 2abee7e18144..d13a1ab8822a 100644
--- a/fs/ceph/debugfs.c
+++ b/fs/ceph/debugfs.c
@@ -170,6 +170,30 @@ static const char * const metric_str[] = {
 	"metadata",
 	"copyfrom"
 };
+
+static int metrics_counters_show(struct seq_file *s, void *p)
+{
+	struct ceph_fs_client *fsc = s->private;
+	struct ceph_client_metric *cm = &fsc->mdsc->metric;
+	u64 count, size_bytes, wait_ns;
+
+	seq_printf(s, "item count size_bytes wait_ns\n");
+
+	for (unsigned i = 0; i < METRIC_MAX; i++) {
+		struct ceph_metric *m = &cm->metric[i];
+		spin_lock(&m->lock);
+		count = m->total;
+		size_bytes = m->size_sum;
+		wait_ns = ktime_to_ns(m->latency_sum);
+		spin_unlock(&m->lock);
+
+		seq_printf(s, "%s %llu %llu %llu\n",
+			   metric_str[i], count, size_bytes, wait_ns);
+	}
+
+	return 0;
+}
+
 static int metrics_latency_show(struct seq_file *s, void *p)
 {
 	struct ceph_fs_client *fsc = s->private;
@@ -368,6 +392,7 @@ DEFINE_SHOW_ATTRIBUTE(caps);
 DEFINE_SHOW_ATTRIBUTE(mds_sessions);
 DEFINE_SHOW_ATTRIBUTE(status);
 DEFINE_SHOW_ATTRIBUTE(metrics_file);
+DEFINE_SHOW_ATTRIBUTE(metrics_counters);
 DEFINE_SHOW_ATTRIBUTE(metrics_latency);
 DEFINE_SHOW_ATTRIBUTE(metrics_size);
 DEFINE_SHOW_ATTRIBUTE(metrics_caps);
@@ -463,6 +488,8 @@ void ceph_fs_debugfs_init(struct ceph_fs_client *fsc)
 
 	debugfs_create_file("file", 0444, fsc->debugfs_metrics_dir, fsc,
 			    &metrics_file_fops);
+	debugfs_create_file("counters", 0444, fsc->debugfs_metrics_dir, fsc,
+			    &metrics_counters_fops);
 	debugfs_create_file("latency", 0444, fsc->debugfs_metrics_dir, fsc,
 			    &metrics_latency_fops);
 	debugfs_create_file("size", 0444, fsc->debugfs_metrics_dir, fsc,
-- 
2.39.2