re-enable the external FTS ICW (#483)

Current ICW external_fts pipeline is not running with the external-fts version.
So change the package and fix the invalid case fts_segment_reset.

Author: jiaqizho

hd-ci/compile_cbdb.bash

@@ -84,7 +84,7 @@ CBDB_VERSION=${CBDB_VERSION}
 EOF
 	else
 		cat <<EOF >>"${SRC_PATH}/cbdb-artifacts.txt"
-external_tar_download_url=${tar_package_path}
+external_tar_download_url="${ROOT_PATH}"/cloudberrydb/gpdb_artifacts/${tar_upload_file_name}
 external_rpm_download_url=${rpm_package_path}
 external_fts_RELEASE_cbdb_${OS_TYPE}_${OS_ARCH}_url=${rpm_download_url}
 external_fts_TARGZ_cbdb_${OS_TYPE}_${OS_ARCH}_url=${tar_download_url}

src/backend/postmaster/interrupt.c

@@ -22,6 +22,7 @@
 #include "storage/latch.h"
 #include "storage/procsignal.h"
 #include "utils/guc.h"
+#include "utils/faultinjector.h"
 
 volatile sig_atomic_t ConfigReloadPending = false;
 volatile sig_atomic_t ShutdownRequestPending = false;
@@ -71,6 +72,8 @@ SignalHandlerForConfigReload(SIGNAL_ARGS)
 void
 SignalHandlerForCrashExit(SIGNAL_ARGS)
 {
+	SIMPLE_FAULT_INJECTOR("fault_in_background_writer_quickdie");
+
 	/*
 	 * We DO NOT want to run proc_exit() or atexit() callbacks -- we're here
 	 * because shared memory may be corrupted, so we don't want to try to

src/test/isolation2/expected/fts_segment_reset.out

@@ -7,10 +7,15 @@
 -- end_matchsubs
 
 -- Let FTS detect/declare failure sooner
-!\retcode gpconfig -c gp_fts_probe_interval -v 10 --masteronly;
-(exited with code 0)
-!\retcode gpstop -u;
-(exited with code 0)
+-- start_ignore
+alter system set gp_fts_probe_interval to 10;
+ALTER
+select pg_reload_conf();
+ pg_reload_conf 
+----------------
+ t              
+(1 row)
+-- end_ignore
 
 -- Let the background writer sleep 27 seconds to delay the resetting.
 -- This number is selected because there's a slight chance that FTS senses
@@ -53,7 +58,8 @@ CREATE
 1<:  <... completed>
 ERROR:  fault triggered, fault name:'start_prepare' fault type:'panic'
 2<:  <... completed>
-CREATE
+DETAIL:  Segments are in reset/recovery mode.
+ERROR:  failed to acquire resources on one or more segments
 
 -- We shouldn't see failover to mirror
 select gp_request_fts_probe_scan();
@@ -79,14 +85,21 @@ select gp_inject_fault('fault_in_background_writer_quickdie', 'reset', dbid) fro
  Success:        
 (1 row)
 
+select pg_sleep(30);
+ pg_sleep 
+----------
+          
+(1 row)
+
+-- start_ignore
+-- restore parameters
+alter system reset gp_fts_probe_interval;
+select pg_reload_conf();
+-- end_ignore
+
 -- The only table that should have been created successfully
 drop table fts_reset_t3;
 DROP
 
 -- In case anything goes wrong, we don't want to affect other tests. So rebalance the cluster anyway.
-!\retcode gprecoverseg -aF !\retcode gprecoverseg -ar 
--- restore parameters
-!\retcode gpconfig -r gp_fts_probe_interval --masteronly;
-(exited with code 127)
-!\retcode gpstop -u;
-(exited with code 0)
+!\retcode gprecoverseg -aF !\retcode gprecoverseg -ar

src/test/isolation2/sql/fts_segment_reset.sql

@@ -7,8 +7,10 @@
 -- end_matchsubs
 
 -- Let FTS detect/declare failure sooner 
-!\retcode gpconfig -c gp_fts_probe_interval -v 10 --masteronly;
-!\retcode gpstop -u;
+-- start_ignore
+alter system set gp_fts_probe_interval to 10;
+select pg_reload_conf();
+-- end_ignore
 
 -- Let the background writer sleep 27 seconds to delay the resetting.
 -- This number is selected because there's a slight chance that FTS senses 
@@ -47,13 +49,17 @@ select dbid, role, preferred_role, status from gp_segment_configuration where co
 select gp_inject_fault('postmaster_server_loop_no_sigkill', 'reset', dbid) from gp_segment_configuration where role = 'p' and content = 0;
 select gp_inject_fault('fault_in_background_writer_quickdie', 'reset', dbid) from gp_segment_configuration where role = 'p' and content = 0;
 
+select pg_sleep(30);
+
+-- start_ignore
+-- restore parameters
+alter system reset gp_fts_probe_interval;
+select pg_reload_conf();
+-- end_ignore
+
 -- The only table that should have been created successfully
 drop table fts_reset_t3;
 
 -- In case anything goes wrong, we don't want to affect other tests. So rebalance the cluster anyway.
 !\retcode gprecoverseg -aF
-!\retcode gprecoverseg -ar
-
--- restore parameters
-!\retcode gpconfig -r gp_fts_probe_interval --masteronly;
-!\retcode gpstop -u;
+!\retcode gprecoverseg -ar