Ceph Configuration Tuning

Modify Ceph configurations to maximize system resource utilization.

You can modify the /etc/ceph/ceph.conf file to configure all Ceph configuration parameters. The configuration procedure is as follows:

1. Go to the /etc/ceph/ directory on the ceph1 node and edit the ceph.conf file.

   1 2	cd /etc/ceph vi ceph.conf

2. Add the following configurations to the file. Table 1 describes the parameters in the configuration file.

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188

   [global] public_network = 192.168.3.0/24 # Adjust the address based on the actual network segment. cluster_network = 192.168.4.0/24 # Adjust the address based on the actual network segment. mon_max_pg_per_osd = 3000 mon_max_pool_pg_num = 300000 ms_bind_before_connect = true ms_dispatch_throttle_bytes = 2097152000 osd_pool_default_min_size = 0 osd_pool_default_pg_num = 1024 osd_pool_default_pgp_num = 1024 osd_pool_default_size = 3 throttler_perf_counter = false bluefs_buffered_io=false osd_max_write_size = 256 osd_enable_op_tracker = false rbd_cache = false [mon] mon_allow_pool_delete = true [osd] rocksdb_cache_index_and_filter_blocks = false rocksdb_cache_size = 2G osd_memory_cache_min = 3G osd_memory_base = 3G bluestore_rocksdb_options = use_direct_reads=true,use_direct_io_for_flush_and_compaction=true,compression=kNoCompression,min_write_buffer_number_to_merge=32,recycle_log_file_num=64,compaction_style=kCompactionStyleLevel,write_buffer_size=64M,target_file_size_base=64M,compaction_threads=32,max_bytes_for_level_multiplier=8,flusher_threads=8,level0_file_num_compaction_trigger=16,level0_slowdown_writes_trigger=36,level0_stop_writes_trigger=48,compaction_readahead_size=524288,max_bytes_for_level_base=536870912,enable_pipelined_write=false,max_background_jobs=16,max_background_flushes=8,max_background_compactions=16,max_write_buffer_number=8,soft_pending_compaction_bytes_limit=137438953472,hard_pending_compaction_bytes_limit=274877906944,delayed_write_rate=33554432 osd_pg_object_context_cache_count = 256 mon_osd_full_ratio = 0.97 mon_osd_nearfull_ratio = 0.95 osd_min_pg_log_entries = 10 osd_max_pg_log_entries = 10 bluestore_cache_meta_ratio = 0.49 bluestore_cache_kv_ratio = 0.49 bluestore_cache_size_ssd = 6G osd_memory_target = 10G bluestore_clone_cow = false [osd.0] osd_numa_node=0 [osd.1] osd_numa_node=0 [osd.2] osd_numa_node=0 [osd.3] osd_numa_node=0 [osd.4] osd_numa_node=0 [osd.5] osd_numa_node=0 [osd.6] osd_numa_node=1 [osd.7] osd_numa_node=1 [osd.8] osd_numa_node=1 [osd.9] osd_numa_node=1 [osd.10] osd_numa_node=1 [osd.11] osd_numa_node=1 [osd.12] osd_numa_node=2 [osd.13] osd_numa_node=2 [osd.14] osd_numa_node=2 [osd.15] osd_numa_node=2 [osd.16] osd_numa_node=2 [osd.17] osd_numa_node=2 [osd.18] osd_numa_node=3 [osd.19] osd_numa_node=3 [osd.20] osd_numa_node=3 [osd.21] osd_numa_node=3 [osd.22] osd_numa_node=3 [osd.23] osd_numa_node=3 [osd.24] osd_numa_node=0 [osd.25] osd_numa_node=0 [osd.26] osd_numa_node=0 [osd.27] osd_numa_node=0 [osd.28] osd_numa_node=0 [osd.29] osd_numa_node=0 [osd.30] osd_numa_node=1 [osd.31] osd_numa_node=1 [osd.32] osd_numa_node=1 [osd.33] osd_numa_node=1 [osd.34] osd_numa_node=1 [osd.35] osd_numa_node=1 [osd.36] osd_numa_node=2 [osd.37] osd_numa_node=2 [osd.38] osd_numa_node=2 [osd.39] osd_numa_node=2 [osd.40] osd_numa_node=2 [osd.41] osd_numa_node=2 [osd.42] osd_numa_node=3 [osd.43] osd_numa_node=3 [osd.44] osd_numa_node=3 [osd.45] osd_numa_node=3 [osd.46] osd_numa_node=3 [osd.47] osd_numa_node=3 [osd.48] osd_numa_node=0 [osd.49] osd_numa_node=0 [osd.50] osd_numa_node=0 [osd.51] osd_numa_node=0 [osd.52] osd_numa_node=0 [osd.53] osd_numa_node=0 [osd.54] osd_numa_node=1 [osd.55] osd_numa_node=1 [osd.56] osd_numa_node=1 [osd.57] osd_numa_node=1 [osd.58] osd_numa_node=1 [osd.59] osd_numa_node=1 [osd.60] osd_numa_node=2 [osd.61] osd_numa_node=2 [osd.62] osd_numa_node=2 [osd.63] osd_numa_node=2 [osd.64] osd_numa_node=2 [osd.65] osd_numa_node=2 [osd.66] osd_numa_node=3 [osd.67] osd_numa_node=3 [osd.68] osd_numa_node=3 [osd.69] osd_numa_node=3 [osd.70] osd_numa_node=3 [osd.71] osd_numa_node=3

   

   In the configuration file, osd_numa_node=0 indicates that the osd0 process is bound to NUMA0. This document uses the Kunpeng 920 7260 processor as an example. This model has four NUMA nodes, and 24 OSDs are deployed on each node. To better utilize hardware performance, you need to evenly bind OSD processes to CPU cores. In this case, every 6 consecutive OSDs are bound to a NUMA node . If the CPU in use has only two NUMA nodes, bind every 12 consecutive OSDs to a NUMA node. Other cases may be processed by analogy. You can run the lscpu command to check the number of CPU NUMA nodes.

   

3. Synchronize the configurations to other nodes.

   1	ceph-deploy --overwrite-conf admin ceph1 ceph2 ceph3

4. Restart OSD processes on all server nodes to validate the configurations.

   1	systemctl restart ceph-osd.target