From efb6b6e868be1be0725125485810ce96e74ed124 Mon Sep 17 00:00:00 2001 From: Alex Auvolat Date: Thu, 26 Jan 2023 15:04:32 +0100 Subject: Disk space report Report available disk space on nodes and calculate cluster-wide available space in `garage stats` (fix #479) --- src/garage/admin.rs | 125 +++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 108 insertions(+), 17 deletions(-) (limited to 'src/garage/admin.rs') diff --git a/src/garage/admin.rs b/src/garage/admin.rs index 305c5c65..4eabebca 100644 --- a/src/garage/admin.rs +++ b/src/garage/admin.rs @@ -15,6 +15,7 @@ use garage_util::time::*; use garage_table::replication::*; use garage_table::*; +use garage_rpc::ring::PARTITION_BITS; use garage_rpc::*; use garage_block::manager::BlockResyncErrorInfo; @@ -783,6 +784,7 @@ impl AdminRpcHandler { for node in ring.layout.node_ids().iter() { let mut opt = opt.clone(); opt.all_nodes = false; + opt.skip_global = true; writeln!(&mut ret, "\n======================").unwrap(); writeln!(&mut ret, "Stats for node {:?}:", node).unwrap(); @@ -799,6 +801,15 @@ impl AdminRpcHandler { Err(e) => writeln!(&mut ret, "Network error: {}", e).unwrap(), } } + + writeln!(&mut ret, "\n======================").unwrap(); + write!( + &mut ret, + "Cluster statistics:\n\n{}", + self.gather_cluster_stats() + ) + .unwrap(); + Ok(AdminRpc::Ok(ret)) } else { Ok(AdminRpc::Ok(self.gather_stats_local(opt)?)) @@ -819,22 +830,6 @@ impl AdminRpcHandler { writeln!(&mut ret, "\nDatabase engine: {}", self.garage.db.engine()).unwrap(); - // Gather ring statistics - let ring = self.garage.system.ring.borrow().clone(); - let mut ring_nodes = HashMap::new(); - for (_i, loc) in ring.partitions().iter() { - for n in ring.get_nodes(loc, ring.replication_factor).iter() { - if !ring_nodes.contains_key(n) { - ring_nodes.insert(*n, 0usize); - } - *ring_nodes.get_mut(n).unwrap() += 1; - } - } - writeln!(&mut ret, "\nRing nodes & partition count:").unwrap(); - for (n, c) in ring_nodes.iter() { - writeln!(&mut ret, " {:?} {}", n, c).unwrap(); - } - // Gather table statistics let mut table = vec![" Table\tItems\tMklItems\tMklTodo\tGcTodo".into()]; table.push(self.gather_table_stats(&self.garage.bucket_table, opt.detailed)?); @@ -881,12 +876,108 @@ impl AdminRpcHandler { .unwrap(); if !opt.detailed { - writeln!(&mut ret, "\nIf values are missing (marked as NC), consider adding the --detailed flag - this will be slow.").unwrap(); + writeln!(&mut ret, "\nIf values are missing above (marked as NC), consider adding the --detailed flag (this will be slow).").unwrap(); + } + + if !opt.skip_global { + write!(&mut ret, "\n{}", self.gather_cluster_stats()).unwrap(); } Ok(ret) } + fn gather_cluster_stats(&self) -> String { + let mut ret = String::new(); + + // Gather storage node and free space statistics + let layout = &self.garage.system.ring.borrow().layout; + let mut node_partition_count = HashMap::::new(); + for short_id in layout.ring_assignation_data.iter() { + let id = layout.node_id_vec[*short_id as usize]; + *node_partition_count.entry(id).or_default() += 1; + } + let node_info = self + .garage + .system + .get_known_nodes() + .into_iter() + .map(|n| (n.id, n)) + .collect::>(); + + let mut table = vec![" ID\tHostname\tZone\tCapacity\tPart.\tDataAvail\tMetaAvail".into()]; + for (id, parts) in node_partition_count.iter() { + let info = node_info.get(id); + let status = info.map(|x| &x.status); + let role = layout.roles.get(id).and_then(|x| x.0.as_ref()); + let hostname = status.map(|x| x.hostname.as_str()).unwrap_or("?"); + let zone = role.map(|x| x.zone.as_str()).unwrap_or("?"); + let capacity = role.map(|x| x.capacity_string()).unwrap_or("?".into()); + let avail_str = |x| match x { + Some((avail, total)) => { + let pct = (avail as f64) / (total as f64) * 100.; + let avail = bytesize::ByteSize::b(avail); + let total = bytesize::ByteSize::b(total); + format!("{}/{} ({:.1}%)", avail, total, pct) + } + None => "?".into(), + }; + let data_avail = avail_str(status.and_then(|x| x.data_disk_avail)); + let meta_avail = avail_str(status.and_then(|x| x.meta_disk_avail)); + table.push(format!( + " {:?}\t{}\t{}\t{}\t{}\t{}\t{}", + id, hostname, zone, capacity, parts, data_avail, meta_avail + )); + } + write!( + &mut ret, + "Storage nodes:\n{}", + format_table_to_string(table) + ) + .unwrap(); + + let meta_part_avail = node_partition_count + .iter() + .filter_map(|(id, parts)| { + node_info + .get(id) + .and_then(|x| x.status.meta_disk_avail) + .map(|c| c.0 / *parts) + }) + .collect::>(); + let data_part_avail = node_partition_count + .iter() + .filter_map(|(id, parts)| { + node_info + .get(id) + .and_then(|x| x.status.data_disk_avail) + .map(|c| c.0 / *parts) + }) + .collect::>(); + if !meta_part_avail.is_empty() && !data_part_avail.is_empty() { + let meta_avail = + bytesize::ByteSize(meta_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS)); + let data_avail = + bytesize::ByteSize(data_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS)); + writeln!( + &mut ret, + "\nEstimated available storage space cluster-wide (might be lower in practice):" + ) + .unwrap(); + if meta_part_avail.len() < node_partition_count.len() + || data_part_avail.len() < node_partition_count.len() + { + writeln!(&mut ret, " data: < {}", data_avail).unwrap(); + writeln!(&mut ret, " metadata: < {}", meta_avail).unwrap(); + writeln!(&mut ret, "A precise estimate could not be given as information is missing for some storage nodes.").unwrap(); + } else { + writeln!(&mut ret, " data: {}", data_avail).unwrap(); + writeln!(&mut ret, " metadata: {}", meta_avail).unwrap(); + } + } + + ret + } + fn gather_table_stats( &self, t: &Arc>, -- cgit v1.2.3