snaphsh tile

Author: cali-jumptrading

book/api/metrics-generated.md

@@ -893,3 +893,15 @@
 | <span class="metrics-name">snapin_&#8203;accounts_&#8203;inserted</span> | gauge | Number of accounts inserted during snpashot loading. Might decrease if snapshot load is aborted and restarted |
 
 </div>
+
+## Snaphs Tile
+
+<div class="metrics">
+
+| Metric | Type | Description |
+|--------|------|-------------|
+| <span class="metrics-name">snaphs_&#8203;state</span> | gauge | State of the tile. 0=hashing, 1=done, 2=shutdown |
+| <span class="metrics-name">snaphs_&#8203;full_&#8203;accounts_&#8203;hashed</span> | gauge | Number of accounts hashed so far from the full snapshot. Might decrease if snapshot load is aborted and restarted |
+| <span class="metrics-name">snaphs_&#8203;incremental_&#8203;accounts_&#8203;hashed</span> | gauge | Number of accounts hashed so far from the incremental snapshot. Might decrease if snapshot load is aborted and restarted |
+
+</div>

src/app/firedancer-dev/commands/backtest.c

@@ -36,6 +36,7 @@ static void
 backtest_topo( config_t * config ) {
   ulong exec_tile_cnt   = config->firedancer.layout.exec_tile_count;
   ulong writer_tile_cnt = config->firedancer.layout.writer_tile_count;
+  ulong hash_tile_cnt   = config->firedancer.layout.hash_tile_count;
 
   fd_topo_t * topo = { fd_topob_new( &config->topo, config->name ) };
   topo->max_page_size = fd_cstr_to_shmem_page_sz( config->hugetlbfs.max_page_size );
@@ -92,13 +93,19 @@ backtest_topo( config_t * config ) {
   fd_topob_wksp( topo, "snaprd" );
   fd_topob_wksp( topo, "snapdc" );
   fd_topob_wksp( topo, "snapin" );
+  fd_topob_wksp( topo, "snaphs" );
   fd_topo_tile_t * snaprd_tile = fd_topob_tile( topo, "snaprd",  "snaprd",  "metric_in",  cpu_idx++, 0, 0 );
   fd_topo_tile_t * snapdc_tile = fd_topob_tile( topo, "snapdc",  "snapdc",  "metric_in",  cpu_idx++, 0, 0 );
   fd_topo_tile_t * snapin_tile = fd_topob_tile( topo, "snapin",  "snapin",  "metric_in",  cpu_idx++, 0, 0 );
   snaprd_tile->allow_shutdown = 1;
   snapdc_tile->allow_shutdown = 1;
   snapin_tile->allow_shutdown = 1;
 
+  for( ulong i=0UL; i<hash_tile_cnt; i++ ) {
+    fd_topo_tile_t * snaphsh_tile = fd_topob_tile( topo, "snaphs", "snaphs", "metric_in", cpu_idx++, 0, 0 );
+    snaphsh_tile->allow_shutdown = 1;
+  }
+
   /**********************************************************************/
   /* Setup backtest->replay link (repair_repla) in topo                 */
   /**********************************************************************/
@@ -121,6 +128,8 @@ backtest_topo( config_t * config ) {
   fd_topob_wksp( topo, "snapdc_rd" );
   fd_topob_wksp( topo, "snapin_rd" );
   fd_topob_wksp( topo, "snap_out" );
+  fd_topob_wksp( topo, "snapin_hsh" );
+  fd_topob_wksp( topo, "snaphsh_out" );
   fd_topob_wksp( topo, "replay_manif" );
   /* TODO: Should be depth of 1 or 2, not 4, but it causes backpressure
      from the replay tile parsing the manifest, remove when this is
@@ -129,8 +138,10 @@ backtest_topo( config_t * config ) {
 
   fd_topob_link( topo, "snap_zstd",   "snap_zstd",   8192UL, 16384UL,    1UL );
   fd_topob_link( topo, "snap_stream", "snap_stream", 2048UL, USHORT_MAX, 1UL );
-  fd_topob_link( topo, "snapdc_rd", "snapdc_rd", 128UL, 0UL, 1UL );
-  fd_topob_link( topo, "snapin_rd", "snapin_rd", 128UL, 0UL, 1UL );
+  fd_topob_link( topo, "snapdc_rd",   "snapdc_rd",   128UL,  0UL,        1UL );
+  fd_topob_link( topo, "snapin_rd",   "snapin_rd",   128UL,  0UL,        1UL );
+  fd_topob_link( topo, "snapin_hsh",  "snapin_hsh",  1024UL, USHORT_MAX, 1UL );
+  FOR(hash_tile_cnt)   fd_topob_link( topo, "snaphsh_out", "snaphsh_out", 512UL, 2048UL, 1UL );
 
   fd_topob_tile_out( topo, "snaprd", 0UL, "snap_zstd",   0UL );
   fd_topob_tile_in ( topo, "snapdc", 0UL, "metric_in",   "snap_zstd", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
@@ -144,6 +155,12 @@ backtest_topo( config_t * config ) {
   fd_topob_tile_in( topo, "snaprd", 0UL, "metric_in", "snapin_rd", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
   fd_topob_tile_out( topo, "snapin", 0UL, "snapin_rd", 0UL );
 
+  fd_topob_tile_out( topo, "snapin", 0UL, "snapin_hsh", 0UL );
+  fd_topob_tile_in(  topo, "snapin", 0UL, "metric_in", "snaphsh_out", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
+
+  FOR(hash_tile_cnt) fd_topob_tile_in(  topo, "snaphs", 0UL, "metric_in",  "snapin_hsh", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
+  FOR(hash_tile_cnt) fd_topob_tile_out( topo, "snaphs", 0UL, "snaphsh_out", 0UL );
+
   /**********************************************************************/
   /* More backtest->replay links in topo                                */
   /**********************************************************************/

src/app/firedancer-dev/main.c

@@ -105,6 +105,7 @@ extern fd_topo_run_tile_t fd_tile_shredcap;
 extern fd_topo_run_tile_t fd_tile_snaprd;
 extern fd_topo_run_tile_t fd_tile_snapdc;
 extern fd_topo_run_tile_t fd_tile_snapin;
+extern fd_topo_run_tile_t fd_tile_snaphs;
 
 fd_topo_run_tile_t * TILES[] = {
   &fd_tile_net,
@@ -149,6 +150,7 @@ fd_topo_run_tile_t * TILES[] = {
   &fd_tile_snaprd,
   &fd_tile_snapdc,
   &fd_tile_snapin,
+  &fd_tile_snaphs,
   NULL,
 };
 

src/app/firedancer/config/default.toml

@@ -688,6 +688,9 @@ user = ""
     # very high TPS rates because the cluster size will be very small.
     shred_tile_count = 1
 
+    # How many snapshot hash tiles to run. 
+    hash_tile_count = 1
+
 # All memory that will be used in Firedancer is pre-allocated in two
 # kinds of pages: huge and gigantic.  Huge pages are 2 MiB and gigantic
 # pages are 1 GiB.  This is done to prevent TLB misses which can have a

src/app/firedancer/topology.c

@@ -200,6 +200,7 @@ fd_topo_initialize( config_t * config ) {
   ulong bank_tile_cnt   = config->layout.bank_tile_count;
   ulong exec_tile_cnt   = config->firedancer.layout.exec_tile_count;
   ulong writer_tile_cnt = config->firedancer.layout.writer_tile_count;
+  ulong hash_tile_cnt   = config->firedancer.layout.hash_tile_count;
   ulong resolv_tile_cnt = config->layout.resolv_tile_count;
 
   int enable_rpc = ( config->rpc.port != 0 );
@@ -286,8 +287,11 @@ fd_topo_initialize( config_t * config ) {
   fd_topob_wksp( topo, "snapdc" );
   fd_topob_wksp( topo, "snaprd" );
   fd_topob_wksp( topo, "snapin" );
+  fd_topob_wksp( topo, "snaphs" );
   fd_topob_wksp( topo, "snapdc_rd" );
   fd_topob_wksp( topo, "snapin_rd" );
+  fd_topob_wksp( topo, "snapin_hsh" );
+  fd_topob_wksp( topo, "snaphsh_out" );
   fd_topob_wksp( topo, "snap_stream" );
   fd_topob_wksp( topo, "snap_zstd" );
   fd_topob_wksp( topo, "snap_out" );
@@ -369,6 +373,9 @@ fd_topo_initialize( config_t * config ) {
   /**/                 fd_topob_link( topo, "snap_out",     "snap_out",     2UL,                                      5UL*(1UL<<30UL),               1UL );
   /**/                 fd_topob_link( topo, "snapdc_rd",    "snapdc_rd",    128UL,                                    0UL,                           1UL );
   /**/                 fd_topob_link( topo, "snapin_rd",    "snapin_rd",    128UL,                                    0UL,                           1UL );
+  /**/                 fd_topob_link( topo, "snapin_hsh",   "snapin_hsh",   1024UL,                                   USHORT_MAX,                    1UL );
+
+  FOR(hash_tile_cnt)   fd_topob_link( topo, "snaphsh_out",  "snaphsh_out",  512UL,                                    2048UL,                        1UL );
 
   /* Replay decoded manifest dcache topo obj */
   fd_topo_obj_t * replay_manifest_dcache = fd_topob_obj( topo, "dcache", "replay_manif" );
@@ -439,6 +446,11 @@ fd_topo_initialize( config_t * config ) {
   fd_topo_tile_t * snapin_tile = fd_topob_tile( topo, "snapin", "snapin", "metric_in", tile_to_cpu[ topo->tile_cnt ], 0, 0 );
   snapin_tile->allow_shutdown = 1;
 
+  for( ulong i=0UL; i<hash_tile_cnt; i++ ) {
+    fd_topo_tile_t * snaphsh_tile = fd_topob_tile( topo, "snaphs", "snaphs", "metric_in", tile_to_cpu[ topo->tile_cnt ], 0, 0 );
+    snaphsh_tile->allow_shutdown = 1;
+  }
+
   /* Database cache */
 
   fd_topo_obj_t * funk_obj = setup_topo_funk( topo, "funk",
@@ -723,6 +735,12 @@ fd_topo_initialize( config_t * config ) {
   fd_topob_tile_in( topo, "snaprd", 0UL, "metric_in", "snapin_rd", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
   fd_topob_tile_out( topo, "snapin", 0UL, "snapin_rd", 0UL );
 
+  fd_topob_tile_out( topo, "snapin",  0UL, "snapin_hsh", 0UL );
+  fd_topob_tile_in(  topo, "snapin", 0UL, "metric_in", "snaphsh_out", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
+
+  FOR(hash_tile_cnt) fd_topob_tile_in( topo,  "snaphs", 0UL, "metric_in", "snapin_hsh", 0UL, FD_TOPOB_RELIABLE, FD_TOPOB_POLLED );
+  FOR(hash_tile_cnt) fd_topob_tile_out( topo, "snaphs", 0UL, "snaphsh_out", 0UL );
+
   if( config->tiles.archiver.enabled ) {
     fd_topob_wksp( topo, "arch_f" );
     fd_topob_wksp( topo, "arch_w" );
@@ -1043,6 +1061,8 @@ fd_topo_configure_tile( fd_topo_tile_t * tile,
 
     } else if( FD_UNLIKELY( !strcmp( tile->name, "snapin" ) ) ) {
       tile->snapin.funk_obj_id            = fd_pod_query_ulong( config->topo.props, "funk",      ULONG_MAX );
+    } else if( FD_UNLIKELY( !strcmp( tile->name, "snaphs" ) ) ) {
+
     } else if( FD_UNLIKELY( !strcmp( tile->name, "arch_f" ) ||
                             !strcmp( tile->name, "arch_w" ) ) ) {
       strncpy( tile->archiver.rocksdb_path, config->tiles.archiver.rocksdb_path, sizeof(tile->archiver.rocksdb_path) );

src/app/shared/fd_config.h

@@ -112,6 +112,7 @@ struct fd_configf {
   struct {
     uint exec_tile_count; /* TODO: redundant ish with bank tile cnt */
     uint writer_tile_count;
+    uint hash_tile_count;
   } layout;
 
   struct {

src/app/shared/fd_config_parse.c

@@ -78,6 +78,7 @@ fd_config_extract_podf( uchar *        pod,
                         fd_configf_t * config ) {
   CFG_POP      ( uint,   layout.exec_tile_count                           );
   CFG_POP      ( uint,   layout.writer_tile_count                         );
+  CFG_POP      ( uint,   layout.hash_tile_count                           );
 
   CFG_POP      ( ulong,  blockstore.shred_max                             );
   CFG_POP      ( ulong,  blockstore.block_max                             );

src/disco/metrics/generate/types.py

@@ -30,6 +30,7 @@ class Tile(Enum):
     SNAPRD = 24
     SNAPDC = 25
     SNAPIN = 26
+    SNAPHS = 27
 
 
 class MetricType(Enum):

src/disco/metrics/generated/fd_metrics_all.c

@@ -59,6 +59,7 @@ const char * FD_METRICS_TILE_KIND_NAMES[FD_METRICS_TILE_KIND_CNT] = {
     "snaprd",
     "snapdc",
     "snapin",
+    "snaphs",
 };
 
 const ulong FD_METRICS_TILE_KIND_SIZES[FD_METRICS_TILE_KIND_CNT] = {
@@ -85,6 +86,7 @@ const ulong FD_METRICS_TILE_KIND_SIZES[FD_METRICS_TILE_KIND_CNT] = {
     FD_METRICS_SNAPRD_TOTAL,
     FD_METRICS_SNAPDC_TOTAL,
     FD_METRICS_SNAPIN_TOTAL,
+    FD_METRICS_SNAPHS_TOTAL,
 };
 const fd_metrics_meta_t * FD_METRICS_TILE_KIND_METRICS[FD_METRICS_TILE_KIND_CNT] = {
     FD_METRICS_NET,
@@ -110,4 +112,5 @@ const fd_metrics_meta_t * FD_METRICS_TILE_KIND_METRICS[FD_METRICS_TILE_KIND_CNT]
     FD_METRICS_SNAPRD,
     FD_METRICS_SNAPDC,
     FD_METRICS_SNAPIN,
+    FD_METRICS_SNAPHS,
 };

src/disco/metrics/generated/fd_metrics_all.h

@@ -24,6 +24,7 @@
 #include "fd_metrics_snaprd.h"
 #include "fd_metrics_snapdc.h"
 #include "fd_metrics_snapin.h"
+#include "fd_metrics_snaphs.h"
 #include "fd_metrics_metric.h"
 /* Start of LINK OUT metrics */
 
@@ -161,7 +162,7 @@ extern const fd_metrics_meta_t FD_METRICS_ALL_LINK_OUT[FD_METRICS_ALL_LINK_OUT_T
 
 #define FD_METRICS_TOTAL_SZ (8UL*253UL)
 
-#define FD_METRICS_TILE_KIND_CNT 23
+#define FD_METRICS_TILE_KIND_CNT 24
 extern const char * FD_METRICS_TILE_KIND_NAMES[FD_METRICS_TILE_KIND_CNT];
 extern const ulong FD_METRICS_TILE_KIND_SIZES[FD_METRICS_TILE_KIND_CNT];
 extern const fd_metrics_meta_t * FD_METRICS_TILE_KIND_METRICS[FD_METRICS_TILE_KIND_CNT];