From 076d06506d77f20de6f8d185698da613d4789b65 Mon Sep 17 00:00:00 2001 From: Rahila Syed Date: Sun, 15 Sep 2024 17:56:06 +0530 Subject: [PATCH] Function to report memory context stats of any backend This function sends a signal to a backend to publish statistics of all its memory contexts. Signal handler sets a flag, which causes the relevant backend to copy its MemoryContextStats to a DSA, as part of next CHECK_FOR_INTERRUPTS(). It there are more that 16MB worth of statistics, the remaining statistics are copied as a cumulative total of the remaining contexts. Once its done, it signals the client backend using a condition variable. The client backend then wakes up, reads the shared memory and returns these values in the form of set of records, one for each memory context, to the user, followed by a cumulative total of the remaining contexts, if any. Each backend and auxiliary process has its own slot for reporting the stats. There is an array of such memory slots of size MaxBackends+NumofAuxiliary processes in fixed shared memory. Each of these slots point to a DSA, which contains the stats to be shared by the corresponding process. Each slot has its own LW lock and condition variable for synchronization and communication between the publishing process and the client backend. --- src/backend/postmaster/autovacuum.c | 4 + src/backend/postmaster/checkpointer.c | 4 + src/backend/postmaster/interrupt.c | 4 + src/backend/postmaster/pgarch.c | 4 + src/backend/postmaster/startup.c | 4 + src/backend/postmaster/walsummarizer.c | 4 + src/backend/storage/ipc/ipci.c | 2 + src/backend/storage/ipc/procsignal.c | 3 + src/backend/tcop/postgres.c | 3 + .../utils/activity/wait_event_names.txt | 1 + src/backend/utils/adt/mcxtfuncs.c | 254 ++++++++++++++- src/backend/utils/init/globals.c | 1 + src/backend/utils/mmgr/mcxt.c | 306 +++++++++++++++++- src/include/catalog/pg_proc.dat | 10 + src/include/miscadmin.h | 1 + src/include/storage/procsignal.h | 1 + src/include/utils/memutils.h | 50 +++ src/test/regress/expected/sysviews.out | 12 + src/test/regress/sql/sysviews.sql | 12 + 19 files changed, 668 insertions(+), 12 deletions(-) diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c index dc3cf87aba..5d01497ada 100644 --- a/src/backend/postmaster/autovacuum.c +++ b/src/backend/postmaster/autovacuum.c @@ -768,6 +768,10 @@ HandleAutoVacLauncherInterrupts(void) /* Process sinval catchup interrupts that happened while sleeping */ ProcessCatchupInterrupt(); + + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); } /* diff --git a/src/backend/postmaster/checkpointer.c b/src/backend/postmaster/checkpointer.c index 982572a75d..9caf8fa018 100644 --- a/src/backend/postmaster/checkpointer.c +++ b/src/backend/postmaster/checkpointer.c @@ -616,6 +616,10 @@ HandleCheckpointerInterrupts(void) /* Perform logging of memory contexts of this process */ if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); } /* diff --git a/src/backend/postmaster/interrupt.c b/src/backend/postmaster/interrupt.c index eedc0980cf..1107ff6d45 100644 --- a/src/backend/postmaster/interrupt.c +++ b/src/backend/postmaster/interrupt.c @@ -48,6 +48,10 @@ HandleMainLoopInterrupts(void) /* Perform logging of memory contexts of this process */ if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); } /* diff --git a/src/backend/postmaster/pgarch.c b/src/backend/postmaster/pgarch.c index 02f91431f5..467a253ccd 100644 --- a/src/backend/postmaster/pgarch.c +++ b/src/backend/postmaster/pgarch.c @@ -865,6 +865,10 @@ HandlePgArchInterrupts(void) if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); + if (ConfigReloadPending) { char *archiveLib = pstrdup(XLogArchiveLibrary); diff --git a/src/backend/postmaster/startup.c b/src/backend/postmaster/startup.c index ef6f98ebcd..17beb8737d 100644 --- a/src/backend/postmaster/startup.c +++ b/src/backend/postmaster/startup.c @@ -192,6 +192,10 @@ HandleStartupProcInterrupts(void) /* Perform logging of memory contexts of this process */ if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); } diff --git a/src/backend/postmaster/walsummarizer.c b/src/backend/postmaster/walsummarizer.c index 48350bec52..b3e6c2b5f0 100644 --- a/src/backend/postmaster/walsummarizer.c +++ b/src/backend/postmaster/walsummarizer.c @@ -876,6 +876,10 @@ HandleWalSummarizerInterrupts(void) /* Perform logging of memory contexts of this process */ if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + + /* Publish memory contexts of this process */ + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); } /* diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c index 7783ba854f..8816ef6903 100644 --- a/src/backend/storage/ipc/ipci.c +++ b/src/backend/storage/ipc/ipci.c @@ -50,6 +50,7 @@ #include "storage/sinvaladt.h" #include "utils/guc.h" #include "utils/injection_point.h" +#include "utils/memutils.h" /* GUCs */ int shared_memory_type = DEFAULT_SHARED_MEMORY_TYPE; @@ -340,6 +341,7 @@ CreateOrAttachShmemStructs(void) StatsShmemInit(); WaitEventCustomShmemInit(); InjectionPointShmemInit(); + MemCtxShmemInit(); } /* diff --git a/src/backend/storage/ipc/procsignal.c b/src/backend/storage/ipc/procsignal.c index 87027f27eb..621726cf03 100644 --- a/src/backend/storage/ipc/procsignal.c +++ b/src/backend/storage/ipc/procsignal.c @@ -688,6 +688,9 @@ procsignal_sigusr1_handler(SIGNAL_ARGS) if (CheckProcSignal(PROCSIG_LOG_MEMORY_CONTEXT)) HandleLogMemoryContextInterrupt(); + if (CheckProcSignal(PROCSIG_GET_MEMORY_CONTEXT)) + HandleGetMemoryContextInterrupt(); + if (CheckProcSignal(PROCSIG_PARALLEL_APPLY_MESSAGE)) HandleParallelApplyMessageInterrupt(); diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c index 184b830168..4436b885d9 100644 --- a/src/backend/tcop/postgres.c +++ b/src/backend/tcop/postgres.c @@ -3500,6 +3500,9 @@ ProcessInterrupts(void) if (LogMemoryContextPending) ProcessLogMemoryContextInterrupt(); + if (PublishMemoryContextPending) + ProcessGetMemoryContextInterrupt(); + if (ParallelApplyMessagePending) HandleParallelApplyMessages(); } diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt index 16144c2b72..7a27b5f680 100644 --- a/src/backend/utils/activity/wait_event_names.txt +++ b/src/backend/utils/activity/wait_event_names.txt @@ -158,6 +158,7 @@ WAL_RECEIVER_EXIT "Waiting for the WAL receiver to exit." WAL_RECEIVER_WAIT_START "Waiting for startup process to send initial data for streaming replication." WAL_SUMMARY_READY "Waiting for a new WAL summary to be generated." XACT_GROUP_UPDATE "Waiting for the group leader to update transaction status at transaction end." +MEM_CTX_PUBLISH "Waiting for backend to publish memory information." ABI_compatibility: diff --git a/src/backend/utils/adt/mcxtfuncs.c b/src/backend/utils/adt/mcxtfuncs.c index 6a6634e1cd..163de2d2d2 100644 --- a/src/backend/utils/adt/mcxtfuncs.c +++ b/src/backend/utils/adt/mcxtfuncs.c @@ -17,28 +17,23 @@ #include "funcapi.h" #include "mb/pg_wchar.h" +#include "miscadmin.h" +#include "access/twophase.h" +#include "nodes/pg_list.h" #include "storage/proc.h" #include "storage/procarray.h" #include "utils/array.h" #include "utils/builtins.h" #include "utils/hsearch.h" +#include "utils/memutils.h" +#include "utils/wait_event_types.h" /* ---------- * The max bytes for showing identifiers of MemoryContext. * ---------- */ -#define MEMORY_CONTEXT_IDENT_DISPLAY_SIZE 1024 -/* - * MemoryContextId - * Used for storage of transient identifiers for - * pg_get_backend_memory_contexts. - */ -typedef struct MemoryContextId -{ - MemoryContext context; - int context_id; -} MemoryContextId; +struct MemoryContextState *memCtxState = NULL; /* * int_list_to_array @@ -305,3 +300,240 @@ pg_log_backend_memory_contexts(PG_FUNCTION_ARGS) PG_RETURN_BOOL(true); } + +/* + * pg_get_remote_backend_memory_contexts + * Signal a backend or an auxiliary process to send its memory contexts. + * + * On receipt of this signal, a backend or an auxiliary process sets the flag + * in the signal handler, which causes the next CHECK_FOR_INTERRUPTS() + * or process-specific interrupt handler to copy the memory context statistics + * in a dynamic shared memory space. The statistics for contexts that do not fit in + * shared memory area are stored as a cumulative total of those contexts, + * at the end in the dynamic shared memory. + * Wait for the backend to send signal on the condition variable after + * writing statistics to a shared memory. + * Once condition variable comes out of sleep, check if the required + * backends statistics are available to read and display. + */ +Datum +pg_get_process_memory_contexts(PG_FUNCTION_ARGS) +{ + int pid = PG_GETARG_INT32(0); + bool get_summary = PG_GETARG_BOOL(1); + PGPROC *proc; + ProcNumber procNumber = INVALID_PROC_NUMBER; + int i; + ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo; + dsa_area *area; + dsa_handle handle; + MemoryContextInfo *memctx_info; + MemoryContext oldContext; + + InitMaterializedSRF(fcinfo, 0); + + /* + * See if the process with given pid is a backend or an auxiliary process. + */ + proc = BackendPidGetProc(pid); + if (proc == NULL) + proc = AuxiliaryPidGetProc(pid); + + /* + * BackendPidGetProc() and AuxiliaryPidGetProc() return NULL if the pid + * isn't valid; but by the time we reach kill(), a process for which we + * get a valid proc here might have terminated on its own. There's no way + * to acquire a lock on an arbitrary process to prevent that. But since + * this mechanism is usually used to debug a backend or an auxiliary + * process running and consuming lots of memory, that it might end on its + * own first and its memory contexts are not logged is not a problem. + */ + if (proc == NULL) + { + /* + * This is just a warning so a loop-through-resultset will not abort + * if one backend terminated on its own during the run. + */ + ereport(WARNING, + (errmsg("PID %d is not a PostgreSQL server process", pid))); + return (Datum) 0; + } + + procNumber = GetNumberFromPGProc(proc); + if (procNumber == MyProcNumber) + { + ereport(WARNING, + (errmsg("cannot return statistics for local backend"), + errhint("Use pg_get_backend_memory_contexts instead"))); + PG_RETURN_NULL(); + } + + /* + * Return statistics of top level 1 and 2 contexts, if get_summary is + * true. + */ + LWLockAcquire(&memCtxState[procNumber].lw_lock, LW_EXCLUSIVE); + memCtxState[procNumber].get_summary = get_summary; + + /* + * Create a DSA segment with maximum size of 16MB, send handle to the + * publishing process for storing the stats. If number of contexts exceed + * 16MB, a cumulative total is stored for such contexts. + */ + if (memCtxState[procNumber].memstats_dsa_handle == DSA_HANDLE_INVALID) + { + oldContext = MemoryContextSwitchTo(TopMemoryContext); + area = dsa_create_ext(memCtxState[procNumber].lw_lock.tranche, DSA_DEFAULT_INIT_SEGMENT_SIZE, + 16 * DSA_DEFAULT_INIT_SEGMENT_SIZE); + MemoryContextSwitchTo(oldContext); + handle = dsa_get_handle(area); + memCtxState[procNumber].memstats_dsa_handle = handle; + /* Pin the mapping so that it doesn't throw a warning */ + dsa_pin(area); + dsa_pin_mapping(area); + } + else + { + area = dsa_attach(memCtxState[procNumber].memstats_dsa_handle); + dsa_pin_mapping(area); + } + LWLockRelease(&memCtxState[procNumber].lw_lock); + if (SendProcSignal(pid, PROCSIG_GET_MEMORY_CONTEXT, procNumber) < 0) + { + ereport(WARNING, + (errmsg("could not send signal to process %d: %m", pid))); + return (Datum) 0; + } + + /* + * Wait for a backend to publish stats, indicated by a valid dsa pointer + * set by the backend. + */ + while (1) + { + LWLockAcquire(&memCtxState[procNumber].lw_lock, LW_EXCLUSIVE); + + /* + * We expect to come out of sleep when the requested process has + * finished publishing the statistics, verified using the a valid dsa + * pointer. + * + * Make sure that the information belongs to pid we requested + * information for, Otherwise loop back and wait for the server + * process to finish publishing statistics. + */ + if (memCtxState[procNumber].proc_id == pid && DsaPointerIsValid(memCtxState[procNumber].memstats_dsa_pointer)) + break; + else + LWLockRelease(&memCtxState[procNumber].lw_lock); + + if (ConditionVariableTimedSleep(&memCtxState[procNumber].memctx_cv, 120000, + WAIT_EVENT_MEM_CTX_PUBLISH)) + { + ereport(WARNING, + (errmsg("Wait for %d process to publish stats timed out, try again", pid))); + if (DsaPointerIsValid(memCtxState[procNumber].memstats_dsa_pointer)) + { + dsa_free(area, memCtxState[procNumber].memstats_dsa_pointer); + memCtxState[procNumber].memstats_dsa_pointer = InvalidDsaPointer; + } + return (Datum) 0; + } + } + if (DsaPointerIsValid(memCtxState[procNumber].memstats_dsa_pointer)) + memctx_info = (MemoryContextInfo *) dsa_get_address(area, memCtxState[procNumber].memstats_dsa_pointer); + /* Backend has finished publishing the stats, read them */ + for (i = 0; i < memCtxState[procNumber].in_memory_stats; i++) + { + ArrayType *path_array; + int path_length; + Datum values[PG_GET_BACKEND_MEMORY_CONTEXTS_COLS]; + bool nulls[PG_GET_BACKEND_MEMORY_CONTEXTS_COLS]; + + memset(values, 0, sizeof(values)); + memset(nulls, 0, sizeof(nulls)); + + if (strlen(memctx_info[i].name) != 0) + values[0] = CStringGetTextDatum(memctx_info[i].name); + else + nulls[0] = true; + if (strlen(memctx_info[i].ident) != 0) + values[1] = CStringGetTextDatum(memctx_info[i].ident); + else + nulls[1] = true; + + values[2] = CStringGetTextDatum(memctx_info[i].type); + path_length = memctx_info[i].path_length; + path_array = construct_array_builtin(memctx_info[i].path, path_length, INT4OID); + values[3] = PointerGetDatum(path_array); + values[4] = Int64GetDatum(memctx_info[i].totalspace); + values[5] = Int64GetDatum(memctx_info[i].nblocks); + values[6] = Int64GetDatum(memctx_info[i].freespace); + values[7] = Int64GetDatum(memctx_info[i].freechunks); + values[8] = Int64GetDatum(memctx_info[i].totalspace - memctx_info[i].freespace); + values[9] = Int32GetDatum(memCtxState[procNumber].proc_id); + + tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls); + } + /* If there are more contexts, display a cumulative total of those */ + if (memCtxState[procNumber].total_stats > i) + { + Datum values[PG_GET_BACKEND_MEMORY_CONTEXTS_COLS]; + bool nulls[PG_GET_BACKEND_MEMORY_CONTEXTS_COLS]; + + values[0] = CStringGetTextDatum(memctx_info[i].name); + nulls[1] = true; + nulls[2] = true; + nulls[3] = true; + values[4] = Int64GetDatum(memctx_info[i].totalspace); + values[5] = Int64GetDatum(memctx_info[i].nblocks); + values[6] = Int64GetDatum(memctx_info[i].freespace); + values[7] = Int64GetDatum(memctx_info[i].freechunks); + values[8] = Int64GetDatum(memctx_info[i].totalspace - memctx_info[i].freespace); + values[9] = Int32GetDatum(memCtxState[procNumber].proc_id); + tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls); + } + + /* DSA free allocation for this client */ + if (DsaPointerIsValid(memCtxState[procNumber].memstats_dsa_pointer)) + { + dsa_free(area, memCtxState[procNumber].memstats_dsa_pointer); + memCtxState[procNumber].memstats_dsa_pointer = InvalidDsaPointer; + } + LWLockRelease(&memCtxState[procNumber].lw_lock); + ConditionVariableCancelSleep(); + dsa_detach(area); + return (Datum) 0; +} + +static Size +MemCtxShmemSize(void) +{ + Size size; + Size TotalProcs = add_size(MaxBackends, add_size(NUM_AUXILIARY_PROCS, max_prepared_xacts)); + + size = TotalProcs * sizeof(MemoryContextState); + return size; +} + +void +MemCtxShmemInit(void) +{ + bool found; + Size TotalProcs = add_size(MaxBackends, add_size(NUM_AUXILIARY_PROCS, max_prepared_xacts)); + + memCtxState = (MemoryContextState *) ShmemInitStruct("MemoryContextState", + MemCtxShmemSize(), + &found); + if (!found) + { + for (int i = 0; i < TotalProcs; i++) + { + ConditionVariableInit(&memCtxState[i].memctx_cv); + LWLockInitialize(&memCtxState[i].lw_lock, LWLockNewTrancheId()); + LWLockRegisterTranche(memCtxState[i].lw_lock.tranche, "mem_context_stats_reporting"); + memCtxState[i].memstats_dsa_handle = DSA_HANDLE_INVALID; + memCtxState[i].memstats_dsa_pointer = InvalidDsaPointer; + } + } +} diff --git a/src/backend/utils/init/globals.c b/src/backend/utils/init/globals.c index 03a54451ac..7fc600ff7b 100644 --- a/src/backend/utils/init/globals.c +++ b/src/backend/utils/init/globals.c @@ -42,6 +42,7 @@ volatile sig_atomic_t IdleStatsUpdateTimeoutPending = false; volatile uint32 InterruptHoldoffCount = 0; volatile uint32 QueryCancelHoldoffCount = 0; volatile uint32 CritSectionCount = 0; +volatile sig_atomic_t PublishMemoryContextPending = false; int MyProcPid; pg_time_t MyStartTime; diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c index bde54326c6..5bdeacb74a 100644 --- a/src/backend/utils/mmgr/mcxt.c +++ b/src/backend/utils/mmgr/mcxt.c @@ -19,16 +19,22 @@ *------------------------------------------------------------------------- */ +#include #include "postgres.h" #include "mb/pg_wchar.h" #include "miscadmin.h" +#include "nodes/pg_list.h" +#include "storage/fd.h" +#include "storage/lwlock.h" +#include "storage/dsm.h" +#include "utils/dsa.h" +#include "utils/hsearch.h" #include "utils/memdebug.h" #include "utils/memutils.h" #include "utils/memutils_internal.h" #include "utils/memutils_memorychunk.h" - static void BogusFree(void *pointer); static void *BogusRealloc(void *pointer, Size size, int flags); static MemoryContext BogusGetChunkContext(void *pointer); @@ -166,6 +172,7 @@ static void MemoryContextStatsInternal(MemoryContext context, int level, static void MemoryContextStatsPrint(MemoryContext context, void *passthru, const char *stats_string, bool print_to_stderr); +static void PublishMemoryContext(MemoryContextInfo * memctx_infos, int curr_id, MemoryContext context, List *path, char *clipped_ident); /* * You should not do memory allocations within a critical section, because @@ -1276,6 +1283,21 @@ HandleLogMemoryContextInterrupt(void) /* latch will be set by procsignal_sigusr1_handler */ } +/* + * HandleGetMemoryContextInterrupt + * Handle receipt of an interrupt indicating publishing of memory + * contexts. + * + * All the actual work is deferred to ProcessLogMemoryContextInterrupt() + */ +void +HandleGetMemoryContextInterrupt(void) +{ + InterruptPending = true; + PublishMemoryContextPending = true; + /* latch will be set by procsignal_sigusr1_handler */ +} + /* * ProcessLogMemoryContextInterrupt * Perform logging of memory contexts of this backend process. @@ -1313,6 +1335,288 @@ ProcessLogMemoryContextInterrupt(void) MemoryContextStatsDetail(TopMemoryContext, 100, 100, false); } +/* + * Run by each backend to publish their memory context + * statistics. It performs a breadth first search + * on the memory context tree, so that the parents + * get a chance to report stats before their children. + * + * Statistics are shared via dynamic shared memory which + * can hold statistics of approx 6700 contexts. Remaining + * contexts statistics is captured as a cumulative total. + */ +void +ProcessGetMemoryContextInterrupt(void) +{ + /* Store the memory context details in shared memory */ + + List *contexts; + + HASHCTL ctl; + HTAB *context_id_lookup; + int context_id = 0; + bool found; + MemoryContext stat_cxt; + MemoryContextInfo *meminfo; + bool get_summary = false; + dsa_area *area; + int num_stats; + int idx = MyProcNumber; + int stats_count = 0; + MemoryContextCounters stat; + + PublishMemoryContextPending = false; + + /* + * The hash table is used for constructing "path" column of + * pg_get_remote_backend_memory_contextis view, similar to its local + * backend couterpart. + */ + + /* + * Make a new context that will contain the hash table, to ease the + * cleanup + */ + + stat_cxt = AllocSetContextCreate(CurrentMemoryContext, + "Memory context statistics", + ALLOCSET_DEFAULT_SIZES); + + ctl.keysize = sizeof(MemoryContext); + ctl.entrysize = sizeof(MemoryContextId); + ctl.hcxt = stat_cxt; + + context_id_lookup = hash_create("pg_get_remote_backend_memory_contexts", + 256, + &ctl, + HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); + + contexts = list_make1(TopMemoryContext); + + /* Compute the number of stats that can fit in the DSM seg */ + num_stats = floor(16 * DSA_DEFAULT_INIT_SEGMENT_SIZE / sizeof(MemoryContextInfo)); + /* Attach to DSA segment */ + LWLockAcquire(&memCtxState[idx].lw_lock, LW_EXCLUSIVE); + area = dsa_attach(memCtxState[idx].memstats_dsa_handle); + memCtxState[idx].proc_id = MyProcPid; + get_summary = memCtxState[idx].get_summary; + + /* + * Traverse the memory context tree to find total number of contexts. If + * summary is requested find the total number of contexts at level 1 and + * 2. + */ + foreach_ptr(MemoryContextData, cur, contexts) + { + MemoryContextId *entry; + List *path = NIL; + + entry = (MemoryContextId *) hash_search(context_id_lookup, &cur, + HASH_ENTER, &found); + entry->context_id = context_id; + + stats_count = stats_count + 1; + /* Append the children of the current context to the main list */ + for (MemoryContext c = cur->firstchild; c != NULL; c = c->nextchild) + contexts = lappend(contexts, c); + + if (!get_summary) + continue; + + /* + * Figure out the transient context_id of this context and each of its + * ancestors. + */ + for (MemoryContext cur_context = cur; cur_context != NULL; cur_context = cur_context->parent) + { + MemoryContextId *cur_entry; + + cur_entry = hash_search(context_id_lookup, &cur_context, HASH_FIND, &found); + + if (!found) + { + elog(LOG, "hash table corrupted, can't construct path value"); + break; + } + path = lcons_int(cur_entry->context_id, path); + } + if (list_length(path) == 3) + { + stats_count = stats_count - 1; + break; + } + } + + /* + * Allocate memory in this process's dsa for storing statistics of the the + * memory contexts upto num_stats, for contexts that don't fit in the DSA + * segment, a cumulative total is written as the last record in the DSA + * segment. + */ + stats_count = stats_count > num_stats ? num_stats : stats_count; + memCtxState[idx].memstats_dsa_pointer = dsa_allocate0(area, stats_count * sizeof(MemoryContextInfo)); + meminfo = (MemoryContextInfo *) dsa_get_address(area, memCtxState[idx].memstats_dsa_pointer); + + foreach_ptr(MemoryContextData, cur, contexts) + { + MemoryContextId *entry; + List *path = NIL; + char clipped_ident[MEMORY_CONTEXT_IDENT_DISPLAY_SIZE]; + + entry = (MemoryContextId *) hash_search(context_id_lookup, &cur, + HASH_ENTER, &found); + entry->context_id = context_id; + + /* + * Figure out the transient context_id of this context and each of its + * ancestors. + */ + for (MemoryContext cur_context = cur; cur_context != NULL; cur_context = cur_context->parent) + { + MemoryContextId *cur_entry; + + cur_entry = hash_search(context_id_lookup, &cur_context, HASH_FIND, &found); + + if (!found) + { + elog(LOG, "hash table corrupted, can't construct path value"); + break; + } + path = lcons_int(cur_entry->context_id, path); + } + /* Trim and copy the identifier if it is not set to NULL */ + if (cur->ident != NULL) + { + int idlen = strlen(cur->ident); + + /* + * Some identifiers such as SQL query string can be very long, + * truncate oversize identifiers. + */ + if (idlen >= MEMORY_CONTEXT_IDENT_DISPLAY_SIZE) + idlen = pg_mbcliplen(cur->ident, idlen, MEMORY_CONTEXT_IDENT_DISPLAY_SIZE - 1); + + memcpy(clipped_ident, cur->ident, idlen); + clipped_ident[idlen] = '\0'; + } + if (context_id <= (num_stats - 2)) + { + /* Copy statistics to DSM memory */ + PublishMemoryContext(meminfo, context_id, cur, path, (cur->ident != NULL ? clipped_ident : NULL)); + } + else + { + /* Examine the context stats */ + memset(&stat, 0, sizeof(stat)); + (*cur->methods->stats) (cur, NULL, NULL, &stat, true); + + meminfo[num_stats - 1].totalspace += stat.totalspace; + meminfo[num_stats - 1].nblocks += stat.nblocks; + meminfo[num_stats - 1].freespace += stat.freespace; + meminfo[num_stats - 1].freechunks += stat.freechunks; + } + /* Display information upto level 2 for summary */ + if (get_summary && list_length(path) == 3) + { + memCtxState[idx].in_memory_stats = context_id; + break; + } + + /* + * DSA max limit is reached, write total of the remaining statistics. + */ + if (context_id == (num_stats - 2) && context_id < (stats_count - 1)) + { + memCtxState[idx].in_memory_stats = context_id + 1; + strncpy(meminfo[num_stats - 1].name, "Remaining Totals", 16); + } + context_id++; + } + if (context_id < (num_stats - 2) && !get_summary) + { + memCtxState[idx].in_memory_stats = context_id; + } + + /* + * Signal the waiting client backend after setting the exit condition flag + */ + memCtxState[idx].total_stats = context_id; + LWLockRelease(&memCtxState[idx].lw_lock); + ConditionVariableBroadcast(&memCtxState[idx].memctx_cv); + /* Delete the hash table memory context */ + MemoryContextDelete(stat_cxt); + + dsa_detach(area); +} + +static void +PublishMemoryContext(MemoryContextInfo * memctx_info, int curr_id, MemoryContext context, List *path, char *clipped_ident) +{ + MemoryContextCounters stat; + char *type; + + if (context->name != NULL) + { + Assert(strlen(context->name) < MEMORY_CONTEXT_IDENT_DISPLAY_SIZE); + strncpy(memctx_info[curr_id].name, context->name, strlen(context->name)); + } + else + memctx_info[curr_id].name[0] = '\0'; + + if (clipped_ident != NULL) + { + /* + * To be consistent with logging output, we label dynahash contexts + * with just the hash table name as with MemoryContextStatsPrint(). + */ + if (!strncmp(context->name, "dynahash", 8)) + { + strncpy(memctx_info[curr_id].name, clipped_ident, strlen(clipped_ident)); + memctx_info[curr_id].ident[0] = '\0'; + } + else + strncpy(memctx_info[curr_id].ident, clipped_ident, strlen(clipped_ident)); + } + else + memctx_info[curr_id].ident[0] = '\0'; + + memctx_info[curr_id].path_length = list_length(path); + foreach_int(i, path) + memctx_info[curr_id].path[foreach_current_index(i)] = Int32GetDatum(i); + + /* Examine the context stats */ + memset(&stat, 0, sizeof(stat)); + (*context->methods->stats) (context, NULL, NULL, &stat, true); + + switch (context->type) + { + case T_AllocSetContext: + type = "AllocSet"; + strncpy(memctx_info[curr_id].type, type, strlen(type)); + break; + case T_GenerationContext: + type = "Generation"; + strncpy(memctx_info[curr_id].type, type, strlen(type)); + break; + case T_SlabContext: + type = "Slab"; + strncpy(memctx_info[curr_id].type, type, strlen(type)); + break; + case T_BumpContext: + type = "Bump"; + strncpy(memctx_info[curr_id].type, type, strlen(type)); + break; + default: + type = "???"; + strncpy(memctx_info[curr_id].type, type, strlen(type)); + break; + } + memctx_info[curr_id].totalspace = stat.totalspace; + memctx_info[curr_id].nblocks = stat.nblocks; + memctx_info[curr_id].freespace = stat.freespace; + memctx_info[curr_id].freechunks = stat.freechunks; +} + void * palloc(Size size) { diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index cbbe8acd38..b205c54710 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -8436,6 +8436,16 @@ prorettype => 'bool', proargtypes => 'int4', prosrc => 'pg_log_backend_memory_contexts' }, +# publishing memory contexts of the specified backend +{ oid => '2173', descr => 'publish memory contexts of the specified backend', + proname => 'pg_get_process_memory_contexts', provolatile => 'v', + prorows => '100', proretset => 't', proparallel => 'r', + prorettype => 'record', proargtypes => 'int4 bool', + proallargtypes => '{int4,bool,text,text,text,_int4,int4,int4,int4,int4,int4,int4}', + proargmodes => '{i,i,o,o,o,o,o,o,o,o,o,o}', + proargnames => '{oid, summary, name, ident, type, path, total_bytes, total_nblocks, free_bytes, free_chunks, used_bytes, pid}', + prosrc => 'pg_get_process_memory_contexts' }, + # non-persistent series generator { oid => '1066', descr => 'non-persistent series generator', proname => 'generate_series', prorows => '1000', diff --git a/src/include/miscadmin.h b/src/include/miscadmin.h index 42a2b38cac..7184727cf1 100644 --- a/src/include/miscadmin.h +++ b/src/include/miscadmin.h @@ -96,6 +96,7 @@ extern PGDLLIMPORT volatile sig_atomic_t IdleSessionTimeoutPending; extern PGDLLIMPORT volatile sig_atomic_t ProcSignalBarrierPending; extern PGDLLIMPORT volatile sig_atomic_t LogMemoryContextPending; extern PGDLLIMPORT volatile sig_atomic_t IdleStatsUpdateTimeoutPending; +extern PGDLLIMPORT volatile sig_atomic_t PublishMemoryContextPending; extern PGDLLIMPORT volatile sig_atomic_t CheckClientConnectionPending; extern PGDLLIMPORT volatile sig_atomic_t ClientConnectionLost; diff --git a/src/include/storage/procsignal.h b/src/include/storage/procsignal.h index 221073def3..8cbf6e201c 100644 --- a/src/include/storage/procsignal.h +++ b/src/include/storage/procsignal.h @@ -35,6 +35,7 @@ typedef enum PROCSIG_WALSND_INIT_STOPPING, /* ask walsenders to prepare for shutdown */ PROCSIG_BARRIER, /* global barrier interrupt */ PROCSIG_LOG_MEMORY_CONTEXT, /* ask backend to log the memory contexts */ + PROCSIG_GET_MEMORY_CONTEXT, /* ask backend to log the memory contexts */ PROCSIG_PARALLEL_APPLY_MESSAGE, /* Message from parallel apply workers */ /* Recovery conflict reasons */ diff --git a/src/include/utils/memutils.h b/src/include/utils/memutils.h index bf93433b78..d1d3b9da93 100644 --- a/src/include/utils/memutils.h +++ b/src/include/utils/memutils.h @@ -18,6 +18,9 @@ #define MEMUTILS_H #include "nodes/memnodes.h" +#include "storage/condition_variable.h" +#include "storage/lmgr.h" +#include "utils/dsa.h" /* @@ -48,7 +51,11 @@ #define AllocHugeSizeIsValid(size) ((Size) (size) <= MaxAllocHugeSize) +#define MEMORY_CONTEXT_IDENT_DISPLAY_SIZE 1024 +#define MEM_CONTEXT_SHMEM_STATS_SIZE 30 +#define MEM_CONTEXT_MAX_LEVEL 64 +#define MAX_TYPE_STRING_LENGTH 64 /* * Standard top-level memory contexts. * @@ -115,6 +122,49 @@ extern MemoryContext AllocSetContextCreateInternal(MemoryContext parent, Size initBlockSize, Size maxBlockSize); +/* Dynamic shared memory state for Memory Context Statistics reporting */ +typedef struct MemoryContextInfo +{ + char name[MEMORY_CONTEXT_IDENT_DISPLAY_SIZE]; + char ident[MEMORY_CONTEXT_IDENT_DISPLAY_SIZE]; + Datum path[MEM_CONTEXT_MAX_LEVEL]; + char type[MAX_TYPE_STRING_LENGTH]; + int path_length; + int64 totalspace; + int64 nblocks; + int64 freespace; + int64 freechunks; +} MemoryContextInfo; + +typedef struct MemoryContextState +{ + ConditionVariable memctx_cv; + LWLock lw_lock; + int proc_id; + int in_memory_stats; + int total_stats; + bool get_summary; + dsa_handle memstats_dsa_handle; + dsa_pointer memstats_dsa_pointer; + +} MemoryContextState; + +/* + * MemoryContextId + * Used for storage of transient identifiers for + * pg_get_backend_memory_contexts. + */ +typedef struct MemoryContextId +{ + MemoryContext context; + int context_id; +} MemoryContextId; + +extern PGDLLIMPORT MemoryContextState * memCtxState; +extern void ProcessGetMemoryContextInterrupt(void); +extern void HandleGetMemoryContextInterrupt(void); +extern void MemCtxShmemInit(void); + /* * This wrapper macro exists to check for non-constant strings used as context * names; that's no longer supported. (Use MemoryContextSetIdentifier if you diff --git a/src/test/regress/expected/sysviews.out b/src/test/regress/expected/sysviews.out index fad7fc3a7e..eecba122c3 100644 --- a/src/test/regress/expected/sysviews.out +++ b/src/test/regress/expected/sysviews.out @@ -222,3 +222,15 @@ select count(distinct utc_offset) >= 24 as ok from pg_timezone_abbrevs; t (1 row) +DO $$ +DECLARE + checkpointer_pid int; + r RECORD; +BEGIN + SELECT pid from pg_stat_activity where backend_type='checkpointer' INTO checkpointer_pid; + + select type, name, ident, total_bytes >= free_bytes + from pg_get_process_memory_contexts(checkpointer_pid, false) where path = '{0}' into r; + RAISE NOTICE '%', r; +END $$; +NOTICE: (AllocSet,TopMemoryContext,,t) diff --git a/src/test/regress/sql/sysviews.sql b/src/test/regress/sql/sysviews.sql index b2a7923754..13a4a0bfe2 100644 --- a/src/test/regress/sql/sysviews.sql +++ b/src/test/regress/sql/sysviews.sql @@ -98,3 +98,15 @@ set timezone_abbreviations = 'Australia'; select count(distinct utc_offset) >= 24 as ok from pg_timezone_abbrevs; set timezone_abbreviations = 'India'; select count(distinct utc_offset) >= 24 as ok from pg_timezone_abbrevs; + +DO $$ +DECLARE + checkpointer_pid int; + r RECORD; +BEGIN + SELECT pid from pg_stat_activity where backend_type='checkpointer' INTO checkpointer_pid; + + select type, name, ident, total_bytes >= free_bytes + from pg_get_process_memory_contexts(checkpointer_pid, false) where path = '{0}' into r; + RAISE NOTICE '%', r; +END $$; -- 2.34.1