From 78f15b7ef1a00e5c83d7128cefaf02f9262b9452 Mon Sep 17 00:00:00 2001 From: Binbin Date: Tue, 21 Mar 2023 23:51:47 +0800 Subject: [PATCH] Fix race in temp rdb delete shutdown test (#11840) I saw this error once, in the FreeBSD Daily CI: ``` *** [err]: Temp rdb will be deleted if we use bg_unlink when shutdown in tests/unit/shutdown.tcl Expected [file exists /xxx/temp-10336.rdb] (context: type eval line 15 cmd {assert {[file exists $temp_rdb]}} proc ::test) ``` The log shows that bgsave was executed, and it was successfully executed in the end: ``` Starting test Temp rdb will be deleted if we use bg_unlink when shutdown in tests/unit/shutdown.tcl 10251:M 22 Feb 2023 11:37:25.441 * Background saving started by pid 10336 10336:C 22 Feb 2023 11:37:27.949 * DB saved on disk 10336:C 22 Feb 2023 11:37:27.949 * Fork CoW for RDB: current 0 MB, peak 0 MB, average 0 MB 10251:M 22 Feb 2023 11:37:28.060 * Background saving terminated with success ``` There may be two reasons: 1. The child process has been created, but it has not created the temp rdb file yet, so [file exists $temp_rdb] check failed. 2. The child process bgsave has been executed successfully and the temp file has been deleted, so [file exists $temp_rdb] check failed. From the logs pint, it should be the case 2, case 1 is too extreme, set rdb-key-save-delay to a higher value to ensure bgsave does not succeed early to avoid this case. --- tests/unit/shutdown.tcl | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/tests/unit/shutdown.tcl b/tests/unit/shutdown.tcl index d0a8ffb6d..b419c83a1 100644 --- a/tests/unit/shutdown.tcl +++ b/tests/unit/shutdown.tcl @@ -3,12 +3,17 @@ start_server {tags {"shutdown external:skip"}} { for {set i 0} {$i < 20} {incr i} { r set $i $i } - # It will cost 2s(20 * 100ms) to dump rdb - r config set rdb-key-save-delay 100000 + r config set rdb-key-save-delay 10000000 # Child is dumping rdb r bgsave - after 100 + wait_for_condition 1000 10 { + [s rdb_bgsave_in_progress] eq 1 + } else { + fail "bgsave did not start in time" + } + after 100 ;# give the child a bit of time for the file to be created + set dir [lindex [r config get dir] 1] set child_pid [get_child_pid 0] set temp_rdb [file join [lindex [r config get dir] 1] temp-${child_pid}.rdb]