3 Commits

Author	SHA1	Message	Date
Heikki Linnakangas	989d78aac8	Buffer the TCP incoming stream on libpq connections. ... Reduces the number of syscalls needed to read the commands from the compute. Here's a snippet of strace output from the pageserver, when performing a sequential scan on a table, with prefetch: 3084934 recvfrom(47, "d", 1, 0, NULL, NULL) = 1 3084934 recvfrom(47, "\0\0\0\37", 4, 0, NULL, NULL) = 4 3084934 recvfrom(47, "\2\1\0\0\0\0\362\302\360\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\0\3", 27, 0, NULL, NULL) = 27 3084934 pread64(28, "\0\0\0\1\0\0\0\0\0\0\0\253 "..., 8192, 25190400) = 8192 3084934 write(45, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\3A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3084934 poll([{fd=46, events=POLLIN}, {fd=48, events=POLLIN}], 2, 60000) = 1 ([{fd=46, revents=POLLIN}]) 3084934 read(46, "\0\0\0\0p\311q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3084934 sendto(47, "d\0\0 \5f\0\0\0\0p\311q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 3084934 recvfrom(47, "d", 1, 0, NULL, NULL) = 1 3084934 recvfrom(47, "\0\0\0\37", 4, 0, NULL, NULL) = 4 3084934 recvfrom(47, "\2\1\0\0\0\0\362\302\360\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\0\4", 27, 0, NULL, NULL) = 27 3084934 pread64(28, " \0=\0L\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0;;\0\0\0\4\4\0"..., 8192, 25198592) = 8192 3084934 write(45, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\4A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3084934 poll([{fd=46, events=POLLIN}, {fd=48, events=POLLIN}], 2, 60000) = 1 ([{fd=46, revents=POLLIN}]) 3084934 read(46, "\0\0\0\0\260\344q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3084934 sendto(47, "d\0\0 \5f\0\0\0\0\260\344q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 3084934 recvfrom(47, "d", 1, 0, NULL, NULL) = 1 3084934 recvfrom(47, "\0\0\0\37", 4, 0, NULL, NULL) = 4 3084934 recvfrom(47, "\2\1\0\0\0\0\362\302\360\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\0\5", 27, 0, NULL, NULL) = 27 3084934 write(45, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\5A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3084934 poll([{fd=46, events=POLLIN}, {fd=48, events=POLLIN}], 2, 60000) = 1 ([{fd=46, revents=POLLIN}]) 3084934 read(46, "\0\0\0\0\330\377q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3084934 sendto(47, "d\0\0 \5f\0\0\0\0\330\377q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 This shows the interaction for three get_page_at_lsn requests. For each request, the pageserver performs three recvfrom syscalls to read the incoming request from the socket. After this patch, those recvfrom calls are gone: 3086123 read(47, "\0\0\0\0\360\222q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3086123 sendto(45, "d\0\0 \5f\0\0\0\0\360\222q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 3086123 pread64(29, " "..., 8192, 25182208) = 8192 3086123 write(46, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\2A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3086123 poll([{fd=47, events=POLLIN}, {fd=49, events=POLLIN}], 2, 60000) = 1 ([{fd=47, revents=POLLIN}]) 3086123 read(47, "\0\0\0\0000\256q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3086123 sendto(45, "d\0\0 \5f\0\0\0\0000\256q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 3086123 pread64(29, "\0\0\0\1\0\0\0\0\0\0\0\253 "..., 8192, 25190400) = 8192 3086123 write(46, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\3A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3086123 poll([{fd=47, events=POLLIN}, {fd=49, events=POLLIN}], 2, 60000) = 1 ([{fd=47, revents=POLLIN}]) 3086123 read(47, "\0\0\0\0p\311q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237\362\0\0\237\362\0"..., 8192) = 8192 3086123 sendto(45, "d\0\0 \5f\0\0\0\0p\311q\1\0\0\4\0\f\1\200\1\0 \4 \0\0\0\0\200\237"..., 8198, MSG_NOSIGNAL, NULL, 0) = 8198 3086123 pread64(29, " \0=\0L\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0;;\0\0\0\4\4\0"..., 8192, 25198592) = 8192 3086123 write(46, "B\0\0\0\25\0\0\0\6\177\0\0002\276\0\0@\f\0\0\0\4A\0\0\32\355\0\0\0\0\1"..., 7010) = 7010 3086123 poll([{fd=47, events=POLLIN}, {fd=49, events=POLLIN}], 2, 60000) = 1 ([{fd=47, revents=POLLIN}]) In this test, the compute sends a batch of prefetch requests, and they are read from the socket in one syscall. That syscall was not captured by the strace snippet above, but there are much fewer of them than before.	2022-10-18 18:46:07 +03:00
Kirill Bulatov	b8eb908a3d	Rename old project name references	2022-09-14 08:14:05 +03:00
Heikki Linnakangas	40c845e57d	Switch to async for all concurrency in the pageserver. ... Instead of spawning helper threads, we now use Tokio tasks. There are multiple Tokio runtimes, for different kinds of tasks. One for serving libpq client connections, another for background operations like GC and compaction, and so on. That's not strictly required, we could use just one runtime, but with this you can still get an overview of what's happening with "top -H". There's one subtle behavior in how TenantState is updated. Before this patch, if you deleted all timelines from a tenant, its GC and compaction loops were stopped, and the tenant went back to Idle state. We no longer do that. The empty tenant stays Active. The changes to test_tenant_tasks.py are related to that. There's still plenty of synchronous code and blocking. For example, we still use blocking std::io functions for all file I/O, and the communication with WAL redo processes is still uses low-level unix poll(). We might want to rewrite those later, but this will do for now. The model is that local file I/O is considered to be fast enough that blocking - and preventing other tasks running in the same thread - is acceptable.	2022-09-12 14:21:00 +03:00