pg_waldump的使用
目錄
pg_waldump作用
pg_waldump可解碼並顯示PostgreSQL預寫日志，用於調試
1、分析pg_waldump出來的wal日志信息
2、wal中的FPI占比很高問題分析
3、full_page_writes作用
4、wal日志大，使用壓縮更靠譜。不建議將full_page_writes設置為false來提升性能，就像把fsync設置為false提升性能一樣是不合理的。
[pg@sdbserver1 pg_wal]$ pg_waldump --help
pg_waldump decodes and displays PostgreSQL write-ahead logs for debugging.
Usage:
pg_waldump [OPTION]... [STARTSEG [ENDSEG]]
Options:
-b, --bkp-details output detailed information about backup blocks 輸出備份塊即full-write-page的詳細信息
-e, --end=RECPTR stop reading at WAL location RECPTR 搜索在此lsn偏移處結束
-f, --follow keep retrying after reaching end of WAL 在到達wal剩余時仍繼續嘗試
-n, --limit=N number of records to display 要顯示的記錄數
-p, --path=PATH directory in which to find log segment files or a --pg_wal的路徑，默認是$PGDATA/pg_wal，也可以指定對應的wal文件
directory with a ./pg_wal that contains such files
(default: current directory, ./pg_wal, $PGDATA/pg_wal)
-r, --rmgr=RMGR only show records generated by resource manager RMGR; 使用--rmgr=list列出有效的資源管理器名稱
use --rmgr=list to list valid resource manager names
-s, --start=RECPTR start reading at WAL location RECPTR 搜索從lsn偏移處開始
-t, --timeline=TLI timeline from which to read log records 從中讀取日志記錄的時間線，
(default: 1 or the value used in STARTSEG)
-V, --version output version information, then exit 輸出版本信息，然后退出
-x, --xid=XID only show records with transaction ID XID 只輸出指定的事務的xlog record
-z, --stats[=record] show statistics instead of records 顯示統計信息而不是記錄
(optionally, show per-record statistics)
-?, --help show this help, then exit --幫助顯示，然后退出
PG10.x+做了一些改動，將關鍵字xlog改成wal
| PG9.x或以下版本 | PG10.x+ |
|---|---|
| pg_xlogdump | pg_waldump |
| pg_xlogfile_name(pg_current_xlog_location()) | pg_walfile_name(pg_current_wal_lsn()) |
| pg_xlog_location_diff(pg_current_xlog_location(),flush_location) | pg_wal_lsn_diff(pg_current_wal_lsn(),flush_lsn) |
| pg_xlog_location_diff(pg_current_xlog_location(),restart_lsn) | pg_wal_lsn_diff(pg_current_wal_lsn(),restart_lsn) |
pg_waldump解析日志
postgres=# create table t_wal(n_id int);
CREATE TABLE
postgres=# select pg_current_wal_lsn();
pg_current_wal_lsn
--------------------
0/8E459DB0
(1 row)
postgres=# insert into t_wal values(1);
INSERT 0 1
postgres=# insert into t_wal values(2);
INSERT 0 1
postgres=# checkpoint;
CHECKPOINT
postgres=# select pg_current_wal_lsn();
pg_current_wal_lsn
--------------------
0/8E459F70
(1 row)
postgres=# \q
[pg@sdbserver1 pg_wal]$ pg_waldump -s 0/8E459DB0 -e 0/8E459F70
rmgr: Heap len (rec/tot): 59/ 59, tx: 7102114, lsn: 0/8E459DB0, prev 0/8E459D78, desc: INSERT+INIT off 1, blkref #0: rel 1663/13287/23713247 blk 0
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7102114, lsn: 0/8E459DF0, prev 0/8E459DB0, desc: COMMIT 2021-06-08 16:49:51.835650 CST
rmgr: Heap len (rec/tot): 59/ 59, tx: 7102115, lsn: 0/8E459E20, prev 0/8E459DF0, desc: INSERT off 2, blkref #0: rel 1663/13287/23713247 blk 0
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7102115, lsn: 0/8E459E60, prev 0/8E459E20, desc: COMMIT 2021-06-08 16:49:55.316314 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/8E459E90, prev 0/8E459E60, desc: RUNNING_XACTS nextXid 7102116 latestCompletedXid 7102115 oldestRunningXid 7102116
rmgr: XLOG len (rec/tot): 106/ 106, tx: 0, lsn: 0/8E459EC8, prev 0/8E459E90, desc: CHECKPOINT_ONLINE redo 0/8E459E90; tli 1; prev tli 1; fpw true; xid 0:7102116; oid 23721430; multi 1; offset 0; oldest xid 563 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 0/0; oldest running xid 7102116; online
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/8E459F38, prev 0/8E459EC8, desc: RUNNING_XACTS nextXid 7102116 latestCompletedXid 7102115 oldestRunningXid 7102116
指定wal文件
postgres=# select pg_current_wal_lsn(),
postgres-# pg_walfile_name(pg_current_wal_lsn()),
postgres-# pg_walfile_name_offset(pg_current_wal_lsn());
pg_current_wal_lsn | pg_walfile_name | pg_walfile_name_offset
--------------------+--------------------------+-------------------------------------
0/FAFEBE50 | 0000000100000000000000FA | (0000000100000000000000FA,16694864)
(1 row)
postgres=# insert into t_wal values(2);
INSERT 0 1
postgres=# checkpoint;
CHECKPOINT
postgres=# select pg_current_wal_lsn(),
pg_walfile_name(pg_current_wal_lsn()),
pg_walfile_name_offset(pg_current_wal_lsn());
pg_current_wal_lsn | pg_walfile_name | pg_walfile_name_offset
--------------------+--------------------------+-------------------------------------
0/FAFEC088 | 0000000100000000000000FA | (0000000100000000000000FA,16695432)
(1 row)
--指定wal文件：0000000100000000000000FA
[pg@sdbserver1 pg_wal]$ pg_waldump $PGDATA/pg_wal/0000000100000000000000FA -s 0/FAFEBE50
rmgr: Heap len (rec/tot): 64/ 268, tx: 7134486, lsn: 0/FAFEBE50, prev 0/FAFEBE18, desc: INSERT off 5, blkref #0: rel 1663/13287/23713247 blk 0 FPW
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7134486, lsn: 0/FAFEBF60, prev 0/FAFEBE50, desc: COMMIT 2021-06-08 18:42:16.028620 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBF90, prev 0/FAFEBF60, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBFC8, prev 0/FAFEBF90, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
rmgr: XLOG len (rec/tot): 106/ 106, tx: 0, lsn: 0/FAFEC018, prev 0/FAFEBFC8, desc: CHECKPOINT_ONLINE redo 0/FAFEBFC8; tli 1; prev tli 1; fpw true; xid 0:7134487; oid 23721430; multi 1; offset 0; oldest xid 563 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 0/0; oldest running xid 7134487; online
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEC088, prev 0/FAFEC018, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
pg_waldump: FATAL: error in WAL record at 0/FAFEC088: invalid record length at 0/FAFEC0C0: wanted 24, got 0
--不指定wal文件，結果是一樣的
[pg@sdbserver1 pg_wal]$ pg_waldump -p $PGDATA/pg_wal -s 0/FAFEBE50
rmgr: Heap len (rec/tot): 64/ 268, tx: 7134486, lsn: 0/FAFEBE50, prev 0/FAFEBE18, desc: INSERT off 5, blkref #0: rel 1663/13287/23713247 blk 0 FPW
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7134486, lsn: 0/FAFEBF60, prev 0/FAFEBE50, desc: COMMIT 2021-06-08 18:42:16.028620 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBF90, prev 0/FAFEBF60, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBFC8, prev 0/FAFEBF90, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
rmgr: XLOG len (rec/tot): 106/ 106, tx: 0, lsn: 0/FAFEC018, prev 0/FAFEBFC8, desc: CHECKPOINT_ONLINE redo 0/FAFEBFC8; tli 1; prev tli 1; fpw true; xid 0:7134487; oid 23721430; multi 1; offset 0; oldest xid 563 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 0/0; oldest running xid 7134487; online
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEC088, prev 0/FAFEC018, desc: RUNNING_XACTS nextXid 7134487 latestCompletedXid 7134486 oldestRunningXid 7134487
pg_waldump: FATAL: error in WAL record at 0/FAFEC088: invalid record length at 0/FAFEC0C0: wanted 24, got 0
各個字段的意思
rmgr: Heap
PostgreSQL內部將WAL日志歸類到20多種不同的資源管理器。這條WAL記錄所屬資源管理器為Heap,即堆表。除了Heap還有Btree,Transaction等。
len (rec/tot): 59/ 59
WAL記錄的總長度是59字節，其中main data部分是59字節(只計數main data可能並不合理，本文的后面會有說明)。
tx: 7102114
事務號
lsn: 0/8E459DB0
本WAL記錄的LSN
prev 0/8E459D78
上條WAL記錄的LSN
desc: INSERT+INIT off 1
這是一條insert類型的記錄(每個資源管理器最多包含16種不同的WAL記錄類型)，tuple在page中的位置為1。
blkref #0: rel 1663/13287/23713247 blk 0
引用的第一個page所屬的對表文件為1663/13287/23713247,塊號為0(即ctid的前半部分)。通過oid2name可以查到是哪個堆表。
[pg@sdbserver1 pg_wal]$ oid2name -o 23713247 -p 6777
From database "postgres":
Filenode Table Name
----------------------
23713247 t_wal
postgres=# select pg_relation_filepath('t_wal');
pg_relation_filepath
----------------------
base/13287/23713247
(1 row)
rmgr=list
[pg@sdbserver1 pg_wal]$ pg_waldump --rmgr=list
XLOG
Transaction
Storage
CLOG
Database
Tablespace
MultiXact
RelMap
Standby
Heap2
Heap
Btree
Hash
Gin
Gist
Sequence
SPGist
BRIN
CommitTs
ReplicationOrigin
Generic
LogicalMessage
使用pgbench數據加壓
[pg@sdbserver1 ~]$ pgbench -i -s 20 pgbenchdb -p 6777
dropping old tables...
注意: 表 "pgbench_accounts" 不存在
注意: 表 "pgbench_branches" 不存在
注意: 表 "pgbench_history" 不存在
注意: 表 "pgbench_tellers" 不存在
creating tables...
generating data...
100000 of 2000000 tuples (5%) done (elapsed 0.44 s, remaining 8.41 s)
200000 of 2000000 tuples (10%) done (elapsed 0.86 s, remaining 7.74 s)
300000 of 2000000 tuples (15%) done (elapsed 2.04 s, remaining 11.54 s)
400000 of 2000000 tuples (20%) done (elapsed 2.36 s, remaining 9.43 s)
500000 of 2000000 tuples (25%) done (elapsed 3.88 s, remaining 11.65 s)
600000 of 2000000 tuples (30%) done (elapsed 5.17 s, remaining 12.06 s)
700000 of 2000000 tuples (35%) done (elapsed 5.48 s, remaining 10.17 s)
800000 of 2000000 tuples (40%) done (elapsed 6.51 s, remaining 9.76 s)
900000 of 2000000 tuples (45%) done (elapsed 7.51 s, remaining 9.18 s)
1000000 of 2000000 tuples (50%) done (elapsed 7.73 s, remaining 7.73 s)
1100000 of 2000000 tuples (55%) done (elapsed 8.78 s, remaining 7.19 s)
1200000 of 2000000 tuples (60%) done (elapsed 9.48 s, remaining 6.32 s)
1300000 of 2000000 tuples (65%) done (elapsed 9.70 s, remaining 5.23 s)
1400000 of 2000000 tuples (70%) done (elapsed 10.50 s, remaining 4.50 s)
1500000 of 2000000 tuples (75%) done (elapsed 13.49 s, remaining 4.50 s)
1600000 of 2000000 tuples (80%) done (elapsed 13.76 s, remaining 3.44 s)
1700000 of 2000000 tuples (85%) done (elapsed 14.87 s, remaining 2.62 s)
1800000 of 2000000 tuples (90%) done (elapsed 15.12 s, remaining 1.68 s)
1900000 of 2000000 tuples (95%) done (elapsed 15.97 s, remaining 0.84 s)
2000000 of 2000000 tuples (100%) done (elapsed 16.72 s, remaining 0.00 s)
vacuuming...
creating primary keys...
done.
[pg@sdbserver1 ~]$ pgbench -r -j2 -c4 -T60 pgbenchdb
connection to database "pgbenchdb" failed:
could not connect to server: Connection refused
Is the server running locally and accepting
connections on Unix domain socket "/tmp/.s.PGSQL.6543"?
[pg@sdbserver1 ~]$ pgbench -r -j2 -c4 -T60 pgbenchdb -p 6777
starting vacuum...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 20
query mode: simple
number of clients: 4
number of threads: 2
duration: 60 s
number of transactions actually processed: 15985
latency average = 15.025 ms
tps = 266.215712 (including connections establishing)
tps = 266.257963 (excluding connections establishing)
statement latencies in milliseconds:
0.006 \set aid random(1, 100000 * :scale)
0.001 \set bid random(1, 1 * :scale)
0.001 \set tid random(1, 10 * :scale)
0.001 \set delta random(-5000, 5000)
0.220 BEGIN;
0.576 UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
0.348 SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
0.464 UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
1.073 UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
0.298 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
11.963 END;
查看wal統計信息
[pg@sdbserver1 pg_wal]$ pg_waldump -p $PGDATA/pg_wal -z -s 0/8E48C2B0 -e 0/9E392748
Type N (%) Record size (%) FPI size (%) Combined size (%)
---- - --- ----------- --- -------- --- ------------- ---
XLOG 5491 ( 2.94) 269059 ( 0.12) 40443352 ( 99.62) 40712411 ( 15.31)
Transaction 16000 ( 8.56) 739416 ( 0.33) 0 ( 0.00) 739416 ( 0.28)
Storage 12 ( 0.01) 504 ( 0.00) 0 ( 0.00) 504 ( 0.00)
CLOG 1 ( 0.00) 30 ( 0.00) 0 ( 0.00) 30 ( 0.00)
Database 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Tablespace 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
MultiXact 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
RelMap 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Standby 34 ( 0.02) 1836 ( 0.00) 0 ( 0.00) 1836 ( 0.00)
Heap2 73844 ( 39.50) 216321593 ( 95.98) 32768 ( 0.08) 216354361 ( 81.34)
Heap 78541 ( 42.01) 7211473 ( 3.20) 41548 ( 0.10) 7253021 ( 2.73)
Btree 13015 ( 6.96) 834561 ( 0.37) 80352 ( 0.20) 914913 ( 0.34)
Hash 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Gin 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Gist 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Sequence 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
SPGist 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
BRIN 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
CommitTs 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
ReplicationOrigin 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
Generic 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
LogicalMessage 0 ( 0.00) 0 ( 0.00) 0 ( 0.00) 0 ( 0.00)
-------- -------- -------- --------
Total 186938 225378472 [84.74%] 40598020 [15.26%] 265976492 [100%]
列名解釋：
`各個列的解釋：``Type：xlog中的日志類型``N： 次數及占比``Record size：記錄的大小及占比``FPI size： full page image大小及占比``combined size： 記錄和FPI的大小總和`
wal日志中FPI占比15.26%，Record占比84.74%，網上搜的是開啟了full_page_writes會導致wal中的FPI占比大，但是我使用pgbench測試了下，差別不大，占比也是15%左右。
分析wal日志主要內容
[pg@sdbserver1 pg_wal]$ pg_waldump -p $PGDATA/pg_wal -s 0/AE2A74E0 -e 0/FAFEAB38 |grep -v " lock "| awk '{print $2" "$16" "$17}'|sort |uniq -c|sort -n -r
182083 Heap2 tuples, blkref
163942 Heap2 xid 7134476
27429 XLOG blkref #0:
88 Heap2 1663/23713278/1249; tid
從wal日志中找到最后一次checkpoint
postgres=# insert into t_wal values(1);
INSERT 0 1
postgres=# checkpoint;
CHECKPOINT
postgres=# \q
--使用pg_waldump解析wal日志
[pg@sdbserver1 pg_wal]$ pg_waldump -p $PGDATA/pg_wal -s 0/FAFEBBD0
rmgr: Heap len (rec/tot): 64/ 196, tx: 7134484, lsn: 0/FAFEBBD0, prev 0/FAFEBB98, desc: INSERT off 3, blkref #0: rel 1663/13287/23713247 blk 0 FPW
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7134484, lsn: 0/FAFEBC98, prev 0/FAFEBBD0, desc: COMMIT 2021-06-08 18:20:55.274801 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBCC8, prev 0/FAFEBC98, desc: RUNNING_XACTS nextXid 7134485 latestCompletedXid 7134484 oldestRunningXid 7134485
rmgr: Heap len (rec/tot): 59/ 59, tx: 7134485, lsn: 0/FAFEBD00, prev 0/FAFEBCC8, desc: INSERT off 4, blkref #0: rel 1663/13287/23713247 blk 0
rmgr: Transaction len (rec/tot): 46/ 46, tx: 7134485, lsn: 0/FAFEBD40, prev 0/FAFEBD00, desc: COMMIT 2021-06-08 18:21:25.749244 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBD70, prev 0/FAFEBD40, desc: RUNNING_XACTS nextXid 7134486 latestCompletedXid 7134485 oldestRunningXid 7134486
rmgr: XLOG len (rec/tot): 106/ 106, tx: 0, lsn: 0/FAFEBDA8, prev 0/FAFEBD70, desc: CHECKPOINT_ONLINE redo 0/FAFEBD70; tli 1; prev tli 1; fpw true; xid 0:7134486; oid 23721430; multi 1; offset 0; oldest xid 563 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 0/0; oldest running xid 7134486; online
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 0/FAFEBE18, prev 0/FAFEBDA8, desc: RUNNING_XACTS nextXid 7134486 latestCompletedXid 7134485 oldestRunningXid 7134486
pg_waldump: FATAL: error in WAL record at 0/FAFEBE18: invalid record length at 0/FAFEBE50: wanted 24, got 0
CHECKPOINT_ONLINE操作的lsn：0/FAFEBDA8，prev上一條lsn是：0/FAFEBD70
pg_controldata
[pg@sdbserver1 pg_wal]$ pg_controldata
pg_control version number: 1100
Catalog version number: 201809051
Database system identifier: 6953123691206530231
Database cluster state: in production
pg_control last modified: Tue 08 Jun 2021 06:21:30 PM CST
Latest checkpoint location: 0/FAFEBDA8
Latest checkpoint's REDO location: 0/FAFEBD70
Latest checkpoint's REDO WAL file: 0000000100000000000000FA
Latest checkpoint's TimeLineID: 1
Latest checkpoint's PrevTimeLineID: 1
Latest checkpoint's full_page_writes: on
Latest checkpoint's NextXID: 0:7134486
Latest checkpoint's NextOID: 23721430
Latest checkpoint's NextMultiXactId: 1
Latest checkpoint's NextMultiOffset: 0
Latest checkpoint's oldestXID: 563
Latest checkpoint's oldestXID's DB: 1
Latest checkpoint's oldestActiveXID: 7134486
Latest checkpoint's oldestMultiXid: 1
Latest checkpoint's oldestMulti's DB: 1
Latest checkpoint's oldestCommitTsXid:0
Latest checkpoint's newestCommitTsXid:0
Time of latest checkpoint: Tue 08 Jun 2021 06:21:30 PM CST
Latest checkpoint location: 0/FAFEBDA8 --和wal日志中解析出得lsn一致
Latest checkpoint's REDO location: 0/FAFEBD70
根據解析wal日志還可以做很多事，比如表數據被誤刪以后，pg_waldump找到誤刪的事務號xid，停止數據庫，然后重置xlog，啟動數據庫后就可以看到被誤刪的數據。
根據pg_waldump找到刪除的事務ID
--創建表
postgres=# create table t_delete(n_id int);
CREATE TABLE
--插入數據
postgres=# insert into t_delete select generate_series(1,10);
INSERT 0 10
--查看當前lsn
postgres=# select pg_current_wal_lsn();
pg_current_wal_lsn
--------------------
2/C001CA88
(1 row)
--刪除數據
postgres=# delete from t_delete where n_id >5;
DELETE 5
--查看刪除后的lsn
postgres=# select pg_current_wal_lsn();
pg_current_wal_lsn
--------------------
2/C001CC38
(1 row)
postgres=# \q
--解析wal日志，事務id為710
[pg@sdbserver2 ~]$ pg_waldump -s 2/C001CA88 -p /opt/pg/abdata/6.0/abase1/pg_wal/
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 2/C001CA88, prev 2/C001CA60, desc: RUNNING_XACTS nextXid 710 latestCompletedXid 709 oldestRunningXid 710
rmgr: Heap len (rec/tot): 54/ 54, tx: 710, lsn: 2/C001CAC0, prev 2/C001CA88, desc: DELETE off 6 KEYS_UPDATED , blkref #0: rel 1663/13231/25679 blk 0
rmgr: Heap len (rec/tot): 54/ 54, tx: 710, lsn: 2/C001CAF8, prev 2/C001CAC0, desc: DELETE off 7 KEYS_UPDATED , blkref #0: rel 1663/13231/25679 blk 0
rmgr: Heap len (rec/tot): 54/ 54, tx: 710, lsn: 2/C001CB30, prev 2/C001CAF8, desc: DELETE off 8 KEYS_UPDATED , blkref #0: rel 1663/13231/25679 blk 0
rmgr: Heap len (rec/tot): 54/ 54, tx: 710, lsn: 2/C001CB68, prev 2/C001CB30, desc: DELETE off 9 KEYS_UPDATED , blkref #0: rel 1663/13231/25679 blk 0
rmgr: Heap len (rec/tot): 54/ 54, tx: 710, lsn: 2/C001CBA0, prev 2/C001CB68, desc: DELETE off 10 KEYS_UPDATED , blkref #0: rel 1663/13231/25679 blk 0
rmgr: Transaction len (rec/tot): 34/ 34, tx: 710, lsn: 2/C001CBD8, prev 2/C001CBA0, desc: COMMIT 2021-06-09 15:30:35.334034 CST
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 2/C001CC00, prev 2/C001CBD8, desc: RUNNING_XACTS nextXid 711 latestCompletedXid 710 oldestRunningXid 711
rmgr: XLOG len (rec/tot): 24/ 24, tx: 0, lsn: 2/C001CC38, prev 2/C001CC00, desc: SWITCH
rmgr: XLOG len (rec/tot): 106/ 106, tx: 0, lsn: 2/C1000028, prev 2/C001CC38, desc: CHECKPOINT_SHUTDOWN redo 2/C1000028; tli 1; prev tli 1; fpw true; xid 0:711; oid 25682; multi 1; offset 0; oldest xid 563 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 0/0; oldest running xid 0; shutdown
rmgr: Standby len (rec/tot): 50/ 50, tx: 0, lsn: 2/C1000098, prev 2/C1000028, desc: RUNNING_XACTS nextXid 711 latestCompletedXid 710 oldestRunningXid 711
pg_waldump: FATAL: error in WAL record at 2/C1000098: invalid record length at 2/C10000D0: wanted 24, got 0
--關閉數據庫
[pg@sdbserver2 ~]$ sh stop_abase1.sh
正在關閉數據庫實例...
waiting for server to shut down......... done
server stopped
--重置事務id為710，有時候設置710可能看不到，可以在設置小一點，710-1
[pg@sdbserver2 ~]$ pg_resetwal -x 710 -D /opt/pg/abdata/6.0/abase1
Write-ahead log reset
--查看表數據，刪除的大於5的數據可以正常看到。
postgres=# select xmin,xmax,* from t_delete;
xmin | xmax | n_id
------+------+------
709 | 0 | 1
709 | 0 | 2
709 | 0 | 3
709 | 0 | 4
709 | 0 | 5
709 | 710 | 6
709 | 710 | 7
709 | 710 | 8
709 | 710 | 9
709 | 710 | 10
(10 rows)
--查看事務id，為710，沒查看一次便會增長一個
postgres=# select txid_current();
txid_current
--------------
710
(1 row)
--再次查看數據，發型又看不到了
postgres=# select xmin,xmax,* from t_delete;
xmin | xmax | n_id
------+------+------
709 | 0 | 1
709 | 0 | 2
709 | 0 | 3
709 | 0 | 4
709 | 0 | 5
(5 rows)
postgres=# select txid_current();
txid_current
--------------
711
(1 row)
postgres=# select xmin,xmax,* from t_delete;
xmin | xmax | n_id
------+------+------
709 | 0 | 1
709 | 0 | 2
709 | 0 | 3
709 | 0 | 4
709 | 0 | 5
(5 rows)
在重置后需要及時的將數據備份出來，可以使用copy，create as等方式將數據備份到其他表。不然隨着xid的增長，數據很快又看不到了。
參考資料：
https://www.postgresql.org/docs/13/pgwaldump.html