libceph: add osd op counter metric support

Message ID	20201110020051.118461-1-xiubli@redhat.com
State	New
Headers	show Return-Path: <ceph-devel-owner@kernel.org> From: xiubli@redhat.com To: idryomov@gmail.com Cc: jlayton@kernel.org, zyan@redhat.com, pdonnell@redhat.com, ceph-devel@vger.kernel.org, Xiubo Li <xiubli@redhat.com> Subject: [PATCH] libceph: add osd op counter metric support Date: Tue, 10 Nov 2020 10:00:51 +0800 Message-Id: <20201110020051.118461-1-xiubli@redhat.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Precedence: bulk
Series	libceph: add osd op counter metric support \| expand libceph: add osd op counter metric support

diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h index 83fa08a06507..9ff9ceed7cb5 100644 --- a/include/linux/ceph/osd_client.h +++ b/include/linux/ceph/osd_client.h @@ -339,6 +339,50 @@ struct ceph_osd_backoff { struct ceph_hobject_id *end; }; +struct ceph_osd_metric { + struct percpu_counter op_ops; + struct percpu_counter op_oplen_avg; + struct percpu_counter op_send; + struct percpu_counter op_send_bytes; + struct percpu_counter op_resend; + struct percpu_counter op_reply; + + struct percpu_counter op_rmw; + struct percpu_counter op_r; + struct percpu_counter op_w; + struct percpu_counter op_pgop; + + struct percpu_counter op_stat; + struct percpu_counter op_create; + struct percpu_counter op_read; + struct percpu_counter op_write; + struct percpu_counter op_writefull; + struct percpu_counter op_append; + struct percpu_counter op_zero; + struct percpu_counter op_truncate; + struct percpu_counter op_delete; + struct percpu_counter op_mapext; + struct percpu_counter op_sparse_read; + struct percpu_counter op_clonerange; + struct percpu_counter op_getxattr; + struct percpu_counter op_setxattr; + struct percpu_counter op_cmpxattr; + struct percpu_counter op_rmxattr; + struct percpu_counter op_resetxattrs; + struct percpu_counter op_call; + struct percpu_counter op_watch; + struct percpu_counter op_notify; + + struct percpu_counter op_omap_rd; + struct percpu_counter op_omap_wr; + struct percpu_counter op_omap_del; + + struct percpu_counter op_linger_active; + struct percpu_counter op_linger_send; + struct percpu_counter op_linger_resend; + struct percpu_counter op_linger_ping; +}; + #define CEPH_LINGER_ID_START 0xffff000000000000ULL struct ceph_osd_client { @@ -371,6 +415,8 @@ struct ceph_osd_client { struct ceph_msgpool msgpool_op; struct ceph_msgpool msgpool_op_reply; + struct ceph_osd_metric metric; + struct workqueue_struct *notify_wq; struct workqueue_struct *completion_wq; }; diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c index 2110439f8a24..b65124cd3fa2 100644 --- a/net/ceph/debugfs.c +++ b/net/ceph/debugfs.c @@ -339,6 +339,54 @@ static void dump_backoffs(struct seq_file *s, struct ceph_osd *osd) mutex_unlock(&osd->lock); } +static void dump_op_metric(struct seq_file *s, struct ceph_osd_client *osdc) +{ + struct ceph_osd_metric *m = &osdc->metric; + + seq_printf(s, " op_ops: %lld\n", percpu_counter_sum(&m->op_ops)); + seq_printf(s, " op_active: %d\n", atomic_read(&osdc->num_requests)); + seq_printf(s, " op_oplen_avg: %lld\n", percpu_counter_sum(&m->op_oplen_avg)); + seq_printf(s, " op_send: %lld\n", percpu_counter_sum(&m->op_send)); + seq_printf(s, " op_send_bytes: %lld\n", percpu_counter_sum(&m->op_send_bytes)); + seq_printf(s, " op_resend: %lld\n", percpu_counter_sum(&m->op_resend)); + seq_printf(s, " op_reply: %lld\n", percpu_counter_sum(&m->op_reply)); + + seq_printf(s, " op_rmw: %lld\n", percpu_counter_sum(&m->op_rmw)); + seq_printf(s, " op_r: %lld\n", percpu_counter_sum(&m->op_r)); + seq_printf(s, " op_w: %lld\n", percpu_counter_sum(&m->op_w)); + seq_printf(s, " op_pgop: %lld\n", percpu_counter_sum(&m->op_pgop)); + + seq_printf(s, " op_stat: %lld\n", percpu_counter_sum(&m->op_stat)); + seq_printf(s, " op_create: %lld\n", percpu_counter_sum(&m->op_create)); + seq_printf(s, " op_read: %lld\n", percpu_counter_sum(&m->op_read)); + seq_printf(s, " op_write: %lld\n", percpu_counter_sum(&m->op_write)); + seq_printf(s, " op_writefull: %lld\n", percpu_counter_sum(&m->op_writefull)); + seq_printf(s, " op_append: %lld\n", percpu_counter_sum(&m->op_append)); + seq_printf(s, " op_zero: %lld\n", percpu_counter_sum(&m->op_zero)); + seq_printf(s, " op_truncate: %lld\n", percpu_counter_sum(&m->op_truncate)); + seq_printf(s, " op_delete: %lld\n", percpu_counter_sum(&m->op_delete)); + seq_printf(s, " op_mapext: %lld\n", percpu_counter_sum(&m->op_mapext)); + seq_printf(s, " op_sparse_read: %lld\n", percpu_counter_sum(&m->op_sparse_read)); + seq_printf(s, " op_clonerange: %lld\n", percpu_counter_sum(&m->op_clonerange)); + seq_printf(s, " op_getxattr: %lld\n", percpu_counter_sum(&m->op_getxattr)); + seq_printf(s, " op_setxattr: %lld\n", percpu_counter_sum(&m->op_setxattr)); + seq_printf(s, " op_cmpxattr: %lld\n", percpu_counter_sum(&m->op_cmpxattr)); + seq_printf(s, " op_rmxattr: %lld\n", percpu_counter_sum(&m->op_rmxattr)); + seq_printf(s, " op_resetxattrs: %lld\n", percpu_counter_sum(&m->op_resetxattrs)); + seq_printf(s, " op_call: %lld\n", percpu_counter_sum(&m->op_call)); + seq_printf(s, " op_watch: %lld\n", percpu_counter_sum(&m->op_watch)); + seq_printf(s, " op_notify: %lld\n", percpu_counter_sum(&m->op_notify)); + + seq_printf(s, " op_omap_rd: %lld\n", percpu_counter_sum(&m->op_omap_rd)); + seq_printf(s, " op_omap_wr: %lld\n", percpu_counter_sum(&m->op_omap_wr)); + seq_printf(s, " op_omap_del: %lld\n", percpu_counter_sum(&m->op_omap_del)); + + seq_printf(s, " op_linger_active: %lld\n", percpu_counter_sum(&m->op_linger_active)); + seq_printf(s, " op_linger_send: %lld\n", percpu_counter_sum(&m->op_linger_send)); + seq_printf(s, " op_linger_resend: %lld\n", percpu_counter_sum(&m->op_linger_resend)); + seq_printf(s, " op_linger_ping: %lld\n", percpu_counter_sum(&m->op_linger_ping)); +} + static int osdc_show(struct seq_file *s, void *pp) { struct ceph_client *client = s->private; @@ -372,6 +420,9 @@ static int osdc_show(struct seq_file *s, void *pp) } up_read(&osdc->lock); + + seq_puts(s, "OP METRIC:\n"); + dump_op_metric(s, osdc); return 0; } diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index 7901ab6c79fd..080bb6f03541 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -2273,6 +2273,8 @@ static void encode_request_finish(struct ceph_msg *msg) static void send_request(struct ceph_osd_request *req) { struct ceph_osd *osd = req->r_osd; + struct ceph_osd_client *osdc = req->r_osdc; + struct ceph_osd_req_op *op; verify_osd_locked(osd); WARN_ON(osd->o_osd != req->r_t.osd); @@ -2309,6 +2311,10 @@ static void send_request(struct ceph_osd_request *req) req->r_sent = osd->o_incarnation; req->r_request->hdr.tid = cpu_to_le64(req->r_tid); ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request)); + + percpu_counter_inc(&osdc->metric.op_send); + for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) + percpu_counter_add(&osdc->metric.op_send_bytes, op->indata_len); } static void maybe_request_map(struct ceph_osd_client *osdc) @@ -2424,6 +2430,116 @@ static void __submit_request(struct ceph_osd_request *req, bool wrlocked) goto again; } +static void osd_acount_op_metric(struct ceph_osd_request *req) +{ + struct ceph_osd_metric *m = &req->r_osdc->metric; + struct ceph_osd_req_op *op; + struct percpu_counter *counter; + + percpu_counter_inc(&m->op_ops); + percpu_counter_add(&m->op_oplen_avg, req->r_num_ops); + + if ((req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_READ)) + == (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_READ)) + percpu_counter_inc(&m->op_rmw); + if (req->r_flags & CEPH_OSD_FLAG_READ) + percpu_counter_inc(&m->op_r); + else if (req->r_flags & CEPH_OSD_FLAG_WRITE) + percpu_counter_inc(&m->op_w); + + if (req->r_flags & CEPH_OSD_FLAG_PGOP) + percpu_counter_inc(&m->op_pgop); + + for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) { + counter = NULL; + + switch (op->op) { + case CEPH_OSD_OP_STAT: + counter = &m->op_stat; + break; + case CEPH_OSD_OP_CREATE: + counter = &m->op_create; + break; + case CEPH_OSD_OP_READ: + counter = &m->op_read; + break; + case CEPH_OSD_OP_WRITE: + counter = &m->op_write; + break; + case CEPH_OSD_OP_WRITEFULL: + counter = &m->op_writefull; + break; + case CEPH_OSD_OP_APPEND: + counter = &m->op_append; + break; + case CEPH_OSD_OP_ZERO: + counter = &m->op_zero; + break; + case CEPH_OSD_OP_TRUNCATE: + counter = &m->op_truncate; + break; + case CEPH_OSD_OP_DELETE: + counter = &m->op_delete; + break; + case CEPH_OSD_OP_MAPEXT: + counter = &m->op_mapext; + break; + case CEPH_OSD_OP_SPARSE_READ: + counter = &m->op_sparse_read; + break; + case CEPH_OSD_OP_GETXATTR: + counter = &m->op_getxattr; + break; + case CEPH_OSD_OP_SETXATTR: + counter = &m->op_setxattr; + break; + case CEPH_OSD_OP_CMPXATTR: + counter = &m->op_cmpxattr; + break; + case CEPH_OSD_OP_RMXATTR: + counter = &m->op_rmxattr; + break; + case CEPH_OSD_OP_RESETXATTRS: + counter = &m->op_resetxattrs; + break; + + // OMAP read operations + case CEPH_OSD_OP_OMAPGETVALS: + case CEPH_OSD_OP_OMAPGETKEYS: + case CEPH_OSD_OP_OMAPGETHEADER: + case CEPH_OSD_OP_OMAPGETVALSBYKEYS: + case CEPH_OSD_OP_OMAP_CMP: + counter = &m->op_omap_rd; + break; + + // OMAP write operations + case CEPH_OSD_OP_OMAPSETVALS: + case CEPH_OSD_OP_OMAPSETHEADER: + counter = &m->op_omap_wr; + break; + + // OMAP del operations + case CEPH_OSD_OP_OMAPCLEAR: + case CEPH_OSD_OP_OMAPRMKEYS: + counter = &m->op_omap_del; + break; + + case CEPH_OSD_OP_CALL: + counter = &m->op_call; + break; + case CEPH_OSD_OP_WATCH: + counter = &m->op_watch; + break; + case CEPH_OSD_OP_NOTIFY: + counter = &m->op_notify; + break; + } + + if (counter) + percpu_counter_inc(counter); + } +} + static void account_request(struct ceph_osd_request *req) { WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK)); @@ -2434,6 +2550,8 @@ static void account_request(struct ceph_osd_request *req) req->r_start_stamp = jiffies; req->r_start_latency = ktime_get(); + + osd_acount_op_metric(req); } static void submit_request(struct ceph_osd_request *req, bool wrlocked) @@ -3156,6 +3274,7 @@ static void send_linger_ping(struct ceph_osd_linger_request *lreq) req->r_tid = atomic64_inc_return(&osdc->last_tid); link_request(lreq->osd, req); send_request(req); + percpu_counter_inc(&osdc->metric.op_linger_ping); } static void linger_submit(struct ceph_osd_linger_request *lreq) @@ -3178,6 +3297,9 @@ static void linger_submit(struct ceph_osd_linger_request *lreq) send_linger(lreq); up_write(&osdc->lock); + + percpu_counter_inc(&osdc->metric.op_linger_send); + percpu_counter_inc(&osdc->metric.op_linger_active); } static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq) @@ -3209,6 +3331,7 @@ static void __linger_cancel(struct ceph_osd_linger_request *lreq) cancel_linger_map_check(lreq); unlink_linger(lreq->osd, lreq); linger_unregister(lreq); + percpu_counter_dec(&lreq->osdc->metric.op_linger_active); } static void linger_cancel(struct ceph_osd_linger_request *lreq) @@ -3768,6 +3891,7 @@ static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg) mutex_unlock(&osd->lock); up_read(&osdc->lock); + percpu_counter_inc(&osdc->metric.op_reply); __complete_request(req); return; @@ -4008,16 +4132,20 @@ static void kick_requests(struct ceph_osd_client *osdc, osd = lookup_create_osd(osdc, req->r_t.osd, true); link_request(osd, req); if (!req->r_linger) { - if (!osd_homeless(osd) && !req->r_t.paused) + if (!osd_homeless(osd) && !req->r_t.paused) { send_request(req); + percpu_counter_inc(&req->r_osdc->metric.op_resend); + } } else { cancel_linger_request(req); } } list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) { - if (!osd_homeless(lreq->osd)) + if (!osd_homeless(lreq->osd)) { send_linger(lreq); + percpu_counter_inc(&lreq->osdc->metric.op_linger_resend); + } list_del_init(&lreq->scan_item); } @@ -4156,8 +4284,10 @@ static void kick_osd_requests(struct ceph_osd *osd) n = rb_next(n); /* cancel_linger_request() */ if (!req->r_linger) { - if (!req->r_t.paused) + if (!req->r_t.paused) { send_request(req); + percpu_counter_inc(&req->r_osdc->metric.op_resend); + } } else { cancel_linger_request(req); } @@ -4167,6 +4297,7 @@ static void kick_osd_requests(struct ceph_osd *osd) rb_entry(n, struct ceph_osd_linger_request, node); send_linger(lreq); + percpu_counter_inc(&lreq->osdc->metric.op_linger_resend); } } @@ -5205,6 +5336,112 @@ void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc) up_write(&osdc->lock); } +static void ceph_metric_destroy(struct ceph_osd_metric *m) +{ + percpu_counter_destroy(&m->op_ops); + percpu_counter_destroy(&m->op_oplen_avg); + percpu_counter_destroy(&m->op_send); + percpu_counter_destroy(&m->op_send_bytes); + percpu_counter_destroy(&m->op_resend); + percpu_counter_destroy(&m->op_reply); + + percpu_counter_destroy(&m->op_rmw); + percpu_counter_destroy(&m->op_r); + percpu_counter_destroy(&m->op_w); + percpu_counter_destroy(&m->op_pgop); + + percpu_counter_destroy(&m->op_stat); + percpu_counter_destroy(&m->op_create); + percpu_counter_destroy(&m->op_read); + percpu_counter_destroy(&m->op_write); + percpu_counter_destroy(&m->op_writefull); + percpu_counter_destroy(&m->op_append); + percpu_counter_destroy(&m->op_zero); + percpu_counter_destroy(&m->op_truncate); + percpu_counter_destroy(&m->op_delete); + percpu_counter_destroy(&m->op_mapext); + percpu_counter_destroy(&m->op_sparse_read); + percpu_counter_destroy(&m->op_clonerange); + percpu_counter_destroy(&m->op_getxattr); + percpu_counter_destroy(&m->op_setxattr); + percpu_counter_destroy(&m->op_cmpxattr); + percpu_counter_destroy(&m->op_rmxattr); + percpu_counter_destroy(&m->op_resetxattrs); + percpu_counter_destroy(&m->op_call); + percpu_counter_destroy(&m->op_watch); + percpu_counter_destroy(&m->op_notify); + + percpu_counter_destroy(&m->op_omap_rd); + percpu_counter_destroy(&m->op_omap_wr); + percpu_counter_destroy(&m->op_omap_del); + + percpu_counter_destroy(&m->op_linger_active); + percpu_counter_destroy(&m->op_linger_send); + percpu_counter_destroy(&m->op_linger_resend); + percpu_counter_destroy(&m->op_linger_ping); +} + +#define CEPH_PERCPU_COUNTER_INIT(counter) \ +do { \ + ret = percpu_counter_init(&m->counter, 0, GFP_NOIO); \ + if (ret) \ + goto err_op_fail; \ +} while (0) + +static int ceph_metric_init(struct ceph_osd_metric *m) +{ + int ret; + + memset(m, 0, sizeof(*m)); + + CEPH_PERCPU_COUNTER_INIT(op_ops); + CEPH_PERCPU_COUNTER_INIT(op_oplen_avg); + CEPH_PERCPU_COUNTER_INIT(op_send); + CEPH_PERCPU_COUNTER_INIT(op_send_bytes); + CEPH_PERCPU_COUNTER_INIT(op_resend); + CEPH_PERCPU_COUNTER_INIT(op_reply); + + CEPH_PERCPU_COUNTER_INIT(op_rmw); + CEPH_PERCPU_COUNTER_INIT(op_r); + CEPH_PERCPU_COUNTER_INIT(op_w); + CEPH_PERCPU_COUNTER_INIT(op_pgop); + + CEPH_PERCPU_COUNTER_INIT(op_stat); + CEPH_PERCPU_COUNTER_INIT(op_create); + CEPH_PERCPU_COUNTER_INIT(op_read); + CEPH_PERCPU_COUNTER_INIT(op_write); + CEPH_PERCPU_COUNTER_INIT(op_writefull); + CEPH_PERCPU_COUNTER_INIT(op_append); + CEPH_PERCPU_COUNTER_INIT(op_zero); + CEPH_PERCPU_COUNTER_INIT(op_truncate); + CEPH_PERCPU_COUNTER_INIT(op_delete); + CEPH_PERCPU_COUNTER_INIT(op_mapext); + CEPH_PERCPU_COUNTER_INIT(op_sparse_read); + CEPH_PERCPU_COUNTER_INIT(op_clonerange); + CEPH_PERCPU_COUNTER_INIT(op_getxattr); + CEPH_PERCPU_COUNTER_INIT(op_setxattr); + CEPH_PERCPU_COUNTER_INIT(op_cmpxattr); + CEPH_PERCPU_COUNTER_INIT(op_rmxattr); + CEPH_PERCPU_COUNTER_INIT(op_resetxattrs); + CEPH_PERCPU_COUNTER_INIT(op_call); + CEPH_PERCPU_COUNTER_INIT(op_watch); + CEPH_PERCPU_COUNTER_INIT(op_notify); + + CEPH_PERCPU_COUNTER_INIT(op_omap_rd); + CEPH_PERCPU_COUNTER_INIT(op_omap_wr); + CEPH_PERCPU_COUNTER_INIT(op_omap_del); + + CEPH_PERCPU_COUNTER_INIT(op_linger_active); + CEPH_PERCPU_COUNTER_INIT(op_linger_send); + CEPH_PERCPU_COUNTER_INIT(op_linger_resend); + CEPH_PERCPU_COUNTER_INIT(op_linger_ping); + return 0; + +err_op_fail: + ceph_metric_destroy(m); + return ret; +} + /* * init, shutdown */ @@ -5257,6 +5494,9 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) if (!osdc->completion_wq) goto out_notify_wq; + if (ceph_metric_init(&osdc->metric) < 0) + goto out_completion_wq; + schedule_delayed_work(&osdc->timeout_work, osdc->client->options->osd_keepalive_timeout); schedule_delayed_work(&osdc->osds_timeout_work, @@ -5264,6 +5504,8 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) return 0; +out_completion_wq: + destroy_workqueue(osdc->completion_wq); out_notify_wq: destroy_workqueue(osdc->notify_wq); out_msgpool_reply: @@ -5302,6 +5544,7 @@ void ceph_osdc_stop(struct ceph_osd_client *osdc) WARN_ON(atomic_read(&osdc->num_requests)); WARN_ON(atomic_read(&osdc->num_homeless)); + ceph_metric_destroy(&osdc->metric); ceph_osdmap_destroy(osdc->osdmap); mempool_destroy(osdc->req_mempool); ceph_msgpool_destroy(&osdc->msgpool_op);

libceph: add osd op counter metric support

Commit Message

Comments

Patch