You are not logged in.
Pages: 1
Let me add my two cents.
For me, there are three things that greatly help in understanding the framework, first two of which I try to do daily:
1. Follow the commits. Honestly, I'm just amazed by Arnaud's productivity throughout the day, especially day after day. Arnaud, please share your secret to such work efficiency!
2. Keep an eye on the discussions forum.
3. When necessary, look at the test units for examples of how to use various functions, approaches, and usage scenarios, as they cover the entire framework.
Incredible. Thank you!
I'll try to do some tests.
Great as always, ab!
Thank you!
Generally speaking also plus FTP, FTPS, and SFTP
ftp://[user[:password]@]server[:port]/path/to/remote/resource.mpeg
sftp://[user[:password]@]server[:port]/path/to/remote/resource.mpeg
I'm afraid to be wrong, but maybe it makes sense to add the ability to download files over SMB to MGet.
Or move this feature into a separate utility, something like
smbget - wget-like utility for download files over SMB
Humm, you are rigth. Testing with "ping" works fine.
But with small test program like
begin
for i := 10 to 80 do
begin
write(IntToStr(i) + ',');
sleep(200);
end;
end.
I get this log
Start at: 07.02.2024 19:54:58
OnRedirect at: 07.02.2024 19:55:06
10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52
OnRedirect at: 07.02.2024 19:55:12
,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,
Exit Code: 0
function TFoo.OnRedirect(const text: RawByteString; pid: cardinal): boolean;
begin
Result := False;
// Writeln(pid);
if text <> '' then
begin
Writeln('OnRedirect at: ' + DateTimeToStr(Now));
Writeln(text);
end;
end;
Thanks ab.
I tested, but the result is the same.
As always, thanks for the lightning-fast response, ab! ))
Add.
Is it possible to add RedirectOutput execution after a certain timeout rather than after the buffer is full?
For long processes with percentage progress, a lot of time passes until the buffer is filled out and the data is transferred to TOnRedirect.
Thanks!
Hi!
mormot.core.os
// - should return true to stop the execution, or false to continue
.....
// - the raw process ID (dword on Windows, cint on POSIX) is also supplied
TOnRedirect = function(const text: RawByteString; pid: cardinal): boolean of object;
// - will optionally call onoutput() to notify the new output state
// - can abort if onoutput() callback returns false, or waitfordelayms expires
....
function RunRedirect(const cmd: TFileName; exitcode: PInteger = nil;
Actually, it works fine as described for TOnRedirect. i.e. when the result is false then execution continue.
Thank you Arnaud for your daily work and for sharing your knowledge and experience with us!
Great job!
Regarding your question in the blog article about the magic number 1.0594631 is the twelfth root of 2, if I'm not mistaken.
And your hint about programming music suggests that there are 12 semitones in an octave and each semitone differs from its neighbors by 1.0594631 (the twelfth root of 2).
I guessed? )))))
Yes, I can create a tutorial.
With some examples of Tasks to do the builds, compilation, etc. This part was where I invested the most time, but it wasn't complicated.
Maybe I missed something. Is there a link to your tutorial? Thank you!
Do you change line 616 cMaxStm = 50 in mormot.db.sql.ibx.
Yes, 48 and below working fine ))). I added the numbers for Ibx firebird 25 to my previous post.
Previous test on 2.5 pass without problem.
Because, as I wrote earlier, embedded version of fbclient.dll from firebird 5 installation was used instead of connecting to 2.5 firebird server
Edited: Or execute statement size exceeded 64K limit on Fb2.5, Fb3 and above have limit of 10MB.
Yes, that's right, https://www.firebirdfaq.org/faq197/
Do you try to add define LEGACYFIREBIRDAPIONLY to force using of legacy api with Fb4/5?
Yes, but without success.
! EXC EIBInterBaseError {Message:"Engine Code: 335544351\r\nunsuccessful metadata update"} [Main] at 0100375a46 client/include/wloadlibrary.inc tfbclientapi.ibdatabaseerror (209)
More accurate test results.
Running tests using Synopse mORMot framework 2.2.6792, compiled with Free Pascal 3.2.2 64 bit, against SQLite 3.44.2, on Windows 11 64bit (10.0.22631), at 2024-01-20 20:02:16.
Insertion speed (rows/second) Read speed (rows/second)
Direct Batch Trans Batch Trans | By one All Virtual All Direct
Zeos firebird 25 2739 76921 5911 58403 | Zeos firebird 25 7811 407398 427313
Zeos firebird 4 2471 57124 7942 41158 | Zeos firebird 4 7868 391282 412609
Zeos firebird 5 1960 70443 7295 49310 | Zeos firebird 5 8212 413479 498529
Ibx firebird 25 756 43030 6628 40509 | Ibx firebird 25 1356 100883 117096
Ibx firebird 4 1824 33112 7650 42821 | Ibx firebird 4 1617 138475 104979
Ibx firebird 5 1367 60364 6479 40248 | Ibx firebird 5 1791 235510 211555
A separate test was run for each driver and each firebird server version.
Ibx firebird 25 did not pass the test due to an error: Solved by changing line 616 cMaxStm = 50 in mormot.db.sql.ibx at cMaxStm = 45
Project PerfTestConsole raised exception class 'EIBInterBaseError' with message:
Engine Code: 335544569
Dynamic SQL Error
-SQL error code = -204
-Implementation limit exceeded
-block size exceeds implementation restriction
In file 'FBClientAPI.pas' at line 412:
2. Correct, this is the reason I create ibx mormot driver. Main goal is to have 0 active transaction if nobody is connected to app/server.
Great, thanks for sharing your experience!
{$define ZEOSTRANS} is only for testing, not recomended for production.
Ок, thanks. Yes, it is clear!
@vs, do you success build with delphi, if yes delphi version please. @ab alearedy commit changes in trunk.
Yes, with my local edits MVCServerFirebirdIbx compiling successful for both (Lazarus and Delphi 12) and working as expected, except FTS.
Tomorrow I'll try to play with ..\ex\extdb-bench\PerfTestConsole.dpr for both (Lazarus and Delphi 12) with the latest ab's commit.
I suspect that the performance figures that I gave a little earlier are not entirely correct.
Most likely the embedded version of fbclient.dll server from firebird 5 installation was used instead of connecting to a particular version of firebird server.
But this is just a guess, I'll try to figure it out tomorrow.
Add.
By the way. Thanks madorin, there is a pretty good library with precise transaction controls for Firebird. Separate for reading and writing transactions, at least.
Maybe this could be a starting point for the next Firebird driver for mORMot 2
https://github.com/madorin/fibplus/
https://github.com/madorin/fibplus/commits/master
ab, thanks for the tip.
Still, there may be a small typo, line 10 in ..\ex\extdb-bench\PerfTestConsole.dpr
{../../src/$R mormot.win.default.manifest.res}
ttomas, thank you for the clarification!
1 and 3 - clear.
2. A few words about this. As I wrote earlier in this thread, this long transaction leads to an increase records versioning for frequently updated tables, which for systems running 24/7 after some time, leads to a dramatically decrease performance for both reading and writing. Firebird 4 solved this problem at some point, but still, a long transaction that opens when the server starts, in my opinion, is not a very correct configuration. Therefore fbintf seems more suitable.
Some test results
Running tests using Synopse mORMot framework 2.2.6749, compiled with Free Pascal 3.2.2 64 bit, against SQLite 3.44.2, on Windows 8 64bit (6.2.9200), at 2024-01-19 13:28:17.
Insertion speed | Read speed
Direct Batch Trans Batch Trans | By one All Virtual All Direct
Zeos firebird 25 2265 60781 16408 65852 | Zeos firebird 25 10884 366770 399840
Zeos firebird 4 2187 55836 6949 63136 | Zeos firebird 4 6944 373538 412864
Zeos firebird 5 2473 65590 8084 52379 | Zeos firebird 5 8154 403177 490196
Ibx firebird 25 2366 43955 15503 40030 | Ibx firebird 25 3850 164076 120290
Ibx firebird 4 1296 37603 3675 49165 | Ibx firebird 4 1798 106503 99975
Ibx firebird 5 1275 30746 4452 40146 | Ibx firebird 5 1967 221028 214850
For some reason log shows Windows 8 64bit, but in fact Windows 11 64bit.
Yes need to be changed to string. I will try to create push request.
Thank you so much.
Fixing these two issues
for Lazarus:
MVCServerFirebirdIbx.dpr(54,17) Error: identifier idents no member "CreateDescendingPK"
MVCServerFirebirdIbx.dpr(95,1) Error: Can't open resource file "D:\Projects\Delphi\Components\Synopse\mORMot2-2.0.stable\ex\mvc-blog\mormot.win.default.manifest.res"
and
for Delphi:
[dcc32 Error] mormot.db.sql.ibx.pas(120): E2010 Incompatible types: 'UTF8String' and 'string'
[dcc32 Error] MVCServerFirebirdIbx.dpr(54): E2003 Undeclared identifier: 'CreateDescendingPK'
MVCServerFirebirdIbx compiling successful for both and working as expected, except FTS
fbintf1.4-0 already installed on Delphi.
For Lazarus:
MVCServerFirebirdIbx.dpr(54,17) Error: identifier idents no member "CreateDescendingPK"
MVCServerFirebirdIbx.dpr(95,1) Error: Can't open resource file "D:\Projects\Delphi\Components\Synopse\mORMot2-2.0.stable\ex\mvc-blog\mormot.win.default.manifest.res"
mormot.db.sql.ibx.pas(119):
property FirebirdLibraryPathName: RawUtf8
but
mormot.db.sql.ibx.pas(87):
fFirebirdLibraryPathName: string;
Two small inaccuracies
commit 2.2.6749
[dcc32 Error] mormot.db.sql.ibx.pas(120): E2010 Incompatible types: 'UTF8String' and 'string'
[dcc32 Error] MVCServerFirebirdIbx.dpr(54): E2003 Undeclared identifier: 'CreateDescendingPK'
For today's commit, "2.2.6617" works fine. Thank you!
But strangely, I followed https://github.com/synopse/mORMot2/commits/master every day and did not see any changes regarding this issue.
Probably I was inattentive and missed it))))
That's why I decided to bother you today.
Hi Ab.
Was the log I sent useful?
Did you try both OU=1\+2 MEDIA and OU=1+2 MEDIA content?
Yes. Same results.
Can you enable the application console, and define a ASNDEBUG conditional and report the output to the console during the search?
Sorry. I do not have permission to publish the log of communication with the corporate AD server. Can I send the log to your email address?
It is MS.
I tried it, but unfortunately parameter passes, but TLdapClient.GetUserInfo returns false.
I.e. Search returns true but SearchResult.Count = 0.
Hi Ab.
The call
TLdapClient.GetGroupInfo('', vBaseDN, vLdapGroup, '', True);
Successfully returns the corresponding group (vLdapGroup) with N members.
One of the member from this group has DN returned by vLdapGroup.member[k] kind of 'CN=Super User,OU=1\+2 MEDIA,OU=Domain Users,DC=1plus2,DC=corp';
To determine sAMAccountName for this user, the next call
TLdapClient.GetUserInfo('', vLdapGroup.member[k], '', vLdapUser);
returns an error
Invalid input name: CN=Super User,OU=1\+2 MEDIA,OU=Domain Users,DC=1plus2,DC=corp
If I remove \ before +, then such a user does not exist.
The actual name of this OU is "1+2 Media" without quotes.
How to correctly call a TLdapClient.GetUserInfo with such a parameter?
In continuation of our conversation, this is not a problem of non ASCII characters, rather this is a problem of special characters for AD.
Thanks a lot.
It looks like firebird 4 solves that issue of commit retaining transaction
I think it is a bad idea to set up an AutoCommit for transaction for a Rest Server that has as backend firebird database.
Below is a snippet from The Firebird Book: A Reference for Database Developers Copyright © 2004 by Helen Borrie and IBPhoenix
COMMIT with the RETAIN Option
The optional RETAIN [SNAPSHOT] extension to the COMMIT statement causes the server to retain a “snapshot” of the physical transaction’s context at the time the statement is executed and start a new transaction as a clone of the committed one. If this so-called soft commit is used on a SNAPSHOT or SNAPSHOT TABLE STABILITY transaction, the cloned transaction preserves the same snapshot of the data as the original transaction had when it started.
Although it does commit the work permanently and thus change the state of the database, COMMIT RETAIN (CommitRetaining) does not release resources. In the lifespan of a logical task that comprises many repetitions of a similar operation, cloning the context reduces some of the overhead that would be incurred by clearing resources each time with COMMIT, only to allocate the identical resources all over again when a new transaction is started. In particular, it preserves the cursors on sets selected and currently “open.”
The same TID remains active in the TSB and never appears there as “committed.” For this reason, it is often referred to as soft commit, in contrast with the “hard” commit performed by an unmodified COMMIT statement. Each soft commit is like a savepoint with no return. Any subsequent ROLLBACK reverses only the changes that have been posted since the last soft commit. The benefit of the soft commit is that it makes life easy for programmers, especially those using components that implement “scrolling dataset” behavior. It was introduced to support the data grid user interface favored by many users of the
Borland Delphi development environment. By retaining the transaction context, the application can display a seamless before-to-after transition that reduces the effort the programmer would otherwise need to invest in starting new transactions, opening new cursors, and resynchronizing them with row sets buffered on the client. Data access implementations frequently combine posting a single update, insert or delete statement with an immediate COMMIT RETAIN in a mechanism that is dubbed “Autocommit.” It is common for interface layers that implement Autocommit capability to “dumb out” explicit control of transactions by starting one invisibly in situations where the programmer-written code attempts to pass a statement without first starting a transaction itself.
Explicit transaction control is worth the extra effort, especially if you are using a connectivity product that exploits the flexible options provided by Firebird. In a busy environment, the COMMIT RETAIN option can save time and resources, but it has some serious disadvantages:
• A snapshot transaction continues to hold the original snapshot in its view, meaning the user does not see the effects of committed changes from other
transactions that were pending at the start of the transaction.
• As long as the same transaction continues being committed with RETAIN, resource “housekeeping” on the server is inhibited, resulting in excessive growth
of memory resources consumed by the TSB. This growth progressively slows down performance, eventually “freezing” the server and, under adverse operating system conditions, even causing it to crash.
• No old record versions made obsolete by operations committed by a COMMIT RETAIN transaction can be garbage collected as long as the original transaction
is never “hard committed.”
@AB
Thank you for the clarification about AutoFree().
@Michael,
Several tests have shown that only in such configuration record versions are not created.
Please, see an sample. vConnection.AutoCommit := False; and vConnection.Commit; After each Upadte/Delete request.
program Project1;
{$APPTYPE CONSOLE}
{$R *.res}
uses
System.SysUtils, ZConnection;
var
vConnection: TZConnection;
i: integer;
begin
vConnection := TZConnection.Create(nil);
try
vConnection.HostName := 'localhost';
vConnection.Database := 'd:\Projects\Tests\REC_VERSION_TEST.FDB';
vConnection.Protocol := 'firebird';
vConnection.User := 'sysdba';
vConnection.Password := 'masterkey';
// vConnection.Properties.Add('hard_commit=true');
// vConnection.AutoCommit := True;
vConnection.AutoCommit := False;
vConnection.Connect;
i := 0;
while True do
begin
inc(i);
if vConnection.ExecuteDirect('UPDATE SIMPLERECORD SET CHANGETIME = CURRENT_TIMESTAMP') then
begin
vConnection.Commit;
Writeln('Udate #' + i.ToString);
end;
sleep(100);
end;
Readln;
finally
vConnection.Free;
end;
end.
When vConnection.AutoCommit := True, value of field *CHANGETIME* is changing.
But, it looks like CommitRetaining not free the transaction environment, so not reset the ID of the oldest active transaction.
@AB
Is it possible at the ORM level to control the transactions behavior?
Thank you!
Please follow the forum rules and don't post huge piece of code or logs.
I'm sorry for that. I will take it into account.
Is 120 lines of code a huge? Can I post a tweaked example here or do I need to save it to an external resource?
Anyway, changing AutoFree () to explicit Create/Free did not solve the problem with record versions.
@AB,
do you recommend not using AutoFree () anymore? (https://synopse.info/forum/viewtopic.ph … 947#p33947)
It was a very useful thing.
Thanks!
After Server restarting
...
Oldest transaction 119
Oldest active 120
Oldest snapshot 120
Next transaction 122
...
SIMPLERECORD (128)
...
Average version length: 9.00, total versions: 113, max versions: 113
So, Oldest active 120 was changed
but total versions: 113 remained the same since there was no read from the table SIMPLERECORD
When the client was started second time
Udate #15
Udate #16
Udate #17
Udate #18
Udate #19
PS>
...
Oldest transaction 120
Oldest active 121
Oldest snapshot 121
Next transaction 142
...
SIMPLERECORD (128)
...
Average version length: 9.00, total versions: 19, max versions: 19
So, sweep has been performed.
But, the problem is that the server keeps an active transaction for the first connection, which gives rise record versions.
@Michael,
@AB
Thanks alot for your help resolving this issue!
But from my point of view issue still exists.
I'm using mORMot 1.18.6286 (mORMot_and_Open_Source_friends_2021-05-06_064801_575871d143) and zeoslib-code-0-r7537-trunk
Source code of test app
program Project1;
{$APPTYPE CONSOLE}
{$R *.res}
uses
System.SysUtils, SynCommons, system.Classes, SynDB, SynTable, ZDbcIntfs, SynDBZeos, mORMot, mORMotDB, mORMotSQLite3,
SynSQLite3Static, mORMotHttpServer, mORMotHttpClient;
type
TSimpleRecord = class(TSQLRecord)
private
FChangeTime: TDateTime;
published
property ChangeTime: TDateTime read FChangeTime write FChangeTime;
end;
var
vCommandLine: TCommandLine;
vMode: RawUTF8;
const
cServerPort = '8080';
cConnectStr = '{"Kind":"TSQLDBZEOSConnectionProperties","ServerName":"firebird://",' +
'"DatabaseName":"d:\\Projects\\Tests\\REC_VERSION_TEST.FDB","User":"SYSDBA","Password":"HuYxjMYsqbe2"}'; //HuYxjMYsqbe2 = masterkey
procedure RunServer;
var
vDBConnectionParams: TSQLDBZEOSConnectionProperties;
vModel: TSQLModel;
vRestServer: TSQLRestServerDB;
vHttpServer: TSQLHttpServer;
vSimpleRecord: TSimpleRecord;
begin
Writeln('RunServer');
TAutoFree.One(vDBConnectionParams, TSQLDBZEOSConnectionProperties.CreateFromJSON(cConnectStr) as TSQLDBZEOSConnectionProperties);
// vDBConnectionParams.ZeosURL.Properties.Add('hard_commit=true');
vModel := TSQLModel.Create([TSimpleRecord]);
VirtualTableExternalRegisterAll(vModel, vDBConnectionParams, True);
TAutoFree.One(vRestServer, TSQLRestServerDB.Create(vModel, ':memory:'));
vRestServer.Model.Owner := vRestServer;
vRestServer.AcquireExecutionMode[execORMGet] := amBackgroundORMSharedThread;
vRestServer.AcquireExecutionMode[execORMWrite] := amBackgroundORMSharedThread;
vRestServer.CreateMissingTables;
if vRestServer.TableRowCount(TSimpleRecord) = 0 then
begin
TAutoFree.One(vSimpleRecord, TSimpleRecord.Create);
vSimpleRecord.ChangeTime := Now;
vRestServer.Add(vSimpleRecord, True);
end;
TAutoFree.One(vHttpServer, TSQLHttpServer.Create(cServerPort, [vRestServer], '+', useHttpApiRegisteringURI));
vHttpServer.AccessControlAllowOrigin := '*';
Writeln('Server is now running on http://localhost:' + cServerPort + '/root'#13#10#13#10+'Press [Enter] to quit');
Readln;
end;
procedure RunClient;
var
i: Integer;
vModel: TSQLModel;
vHttpClient: TSQLHttpClient;
vSimpleRecord: TSimpleRecord;
begin
Writeln('RunClient');
vModel := TSQLModel.Create([TSimpleRecord]);
TAutoFree.One(vHttpClient, TSQLHttpClient.Create('localhost', cServerPort, vModel));
vHttpClient.Model.Owner := vHttpClient;
TAutoFree.One(vSimpleRecord, TSimpleRecord.Create);
vHttpClient.Retrieve(1, vSimpleRecord);
i := 0;
while true do
begin
inc(i);
vSimpleRecord.FChangeTime := now;
if vHttpClient.Update(vSimpleRecord) then
Writeln('Udate #' + i.ToString);
Sleep(100);
if ConsoleKeyPressed(13) then break;
end;
Readln;
end;
begin
TAutoFree.One(vCommandLine, TCommandLine.Create);
vMode := vCommandLine.AsUTF8('mode', 'S', '');
case vMode[1] of
'S', 's': RunServer;
'C', 'c': RunClient;
end;
end.
Output of Project1.exe -mode C
....
Udate #96
Udate #97
Udate #98
Udate #99
Udate #100
Udate #101
Udate #102
Udate #103
Udate #104
Udate #105
Udate #106
Udate #107
Udate #108
Udate #109
Udate #110
Udate #111
Udate #112
Udate #113
PS>
Output of gstat -u sysdba -p masterkey -r d:\Projects\Tests\REC_VERSION_TEST.FDB
$ gstat -u sysdba -p masterkey -r d:\Projects\Tests\REC_VERSION_TEST.FDB
Database "d:\Projects\Tests\REC_VERSION_TEST.FDB"
Database header page information:
Flags 0
Checksum 12345
Generation 120
Page size 4096
ODS version 11.2
Oldest transaction 1
Oldest active 1
Oldest snapshot 1
Next transaction 118
Bumped transaction 1
Sequence number 0
Next attachment ID 1
Implementation ID 16
Shadow count 0
Page buffers 0
Next header page 0
Database dialect 3
Creation date May 10, 2021 23:51:00
Attributes force write
Variable header data:
*END*
Database file sequence:
File d:\Projects\Tests\REC_VERSION_TEST.FDB is the only file
Analyzing database pages ...
SIMPLERECORD (128)
Primary pointer page: 168, Index root page: 169
Average record length: 17.00, total records: 1
Average version length: 9.00, total versions: 113, max versions: 113
Data pages: 1, data page slots: 1, average fill: 73%
Fill distribution:
0 - 19% = 0
20 - 39% = 0
40 - 59% = 0
60 - 79% = 1
80 - 99% = 0
Index NDXSIMPLERECORDID (1)
Depth: 1, leaf buckets: 1, nodes: 1
Average data length: 9.00, total dup: 0, max dup: 0
Fill distribution:
0 - 19% = 1
20 - 39% = 0
40 - 59% = 0
60 - 79% = 0
80 - 99% = 0
Index RDB$PRIMARY1 (0)
Depth: 1, leaf buckets: 1, nodes: 1
Average data length: 9.00, total dup: 0, max dup: 0
Fill distribution:
0 - 19% = 1
20 - 39% = 0
40 - 59% = 0
60 - 79% = 0
80 - 99% = 0
Note for this value, please.
...
Oldest transaction 1
Oldest active 1
Oldest snapshot 1
Next transaction 118
...
SIMPLERECORD (128)
...
Average version length: 9.00, total versions: 113, max versions: 113
Client was closed.
Server still running and hold
Oldest active 1
So, I have
total versions: 113, max versions: 113 equal to the number of updates to the record in the SIMPLERECORD table
Maybe I'm missing something?
Thanks!
I found a solution. Switching to the FireDAC library.
If I try this code,
while true do
begin
inc(i);
vSQLRecordSample.LastLogin := now;
if FRestServer.Update(vSQLRecordSample) then
Writeln('Udate #' + i.ToString);
Sleep(100);
if ConsoleKeyPressed(13) then break;
end;
the number of record versions is 1!
I suppose the automatic transaction used within a BATCH process works as expected, right?
I tried code like this
while true do
begin
inc(i);
vBatch := TSQLRestBatch.Create(FRestServer, TSQLRecordSample, 5);
try
vSQLRecordSample.LastLogin := now;
vBatch.Update(vSQLRecordSample);
vBatch.Update(vSQLRecordSample);
vBatch.Update(vSQLRecordSample);
vBatch.Update(vSQLRecordSample);
vBatch.Update(vSQLRecordSample);
FRestServer.BatchSend(vBatch);
finally
vBatch.Free;
end;
Writeln('Udate #' + i.ToString);
Sleep(300);
if ConsoleKeyPressed(13) then break;
end;
Result remains the same. Amount of record versions is growing at each FRestServer.BatchSend
Hi Michael!
Thanks for your help.
I can't see a zeos regression. Hope you use minimum version v7.2.10 or v8.0 from trunk.
Yes, I am using version 7.2.10-stable build at 2021-01-12 08:55:31
If you have such "old" transactions than you have open cursors somewhere propably
There is no explicit cursor control in my code.
Take care you read all results until no more row is returned. Another cause of many ransactions might be nested transactions with multiple connections
Only one record is read and only one client connection is established.
My test has a server and one client that executes code like this
TAutoFree.One(vMySQLRecord, TMySQLRecord.Create);
FHttpClient.Retrieve(1, vMySQLRecord);
i := 0;
while true do
begin
inc(i);
vMySQLRecord.LastCheckTime := now;
FHttpClient.Update(vMySQLRecord);
Writeln('Udate #' + i.ToString);
Sleep(300);
end;
Actually there are two connections to the database.
First from the main server thread and second from the ThreadSafeConnection, which occurs when the client is selecting data for vMySQLRecord (FHttpClient.Retrieve(1, vMySQLRecord);)
Moreover, if after some time on the server side execute
FDBConnectionParams.ThreadSafeConnection.Disconnect;
For some reason first connections will close and client continues to successfully update vMySQLRecord through the second connection.
And the saddest thing, the ID of the oldest active transaction is equal to the ID of first transaction of the second connection. This is the reason lot of record versions.
Note a second call to IZConnection.StartTransaction creates a savepoint. First call to IZConnection.Commit/Rollback just releases/rollback the savepoint but not the transaction you wanna close.
This is interesting, but the data in the database are updated, therefore commit occurs and savepoint is released.
According the "hard_commit" option: If you are adding that option after creating a transaction then the option is a NOOP.
FDBConnectionParams.ZeosURL.Properties.Add('hard_commit=true')
This code is executed before the call
//SynDBZeos
constructor TSQLDBZEOSConnection.Create(aProperties: TSQLDBConnectionProperties);
var url: TZURL;
begin
inherited Create(aProperties);
url := (fProperties as TSQLDBZEOSConnectionProperties).fURL;
fDatabase := DriverManager.GetConnectionWithParams(url.URL,url.Properties);
// EG: setup the connection transaction behavior now, not once Opened in Connect
//fDatabase.SetReadOnly(false); // is default
// about transactions, see https://synopse.info/forum/viewtopic.php?id=2209
//fDatabase.SetAutoCommit(true); // is default
fDatabase.SetTransactionIsolation(tiReadCommitted); // will be swapped to tiSerialiable for SQLite
end;
In this case url.Properties.Commatext = 'codepage=UTF8,hard_commit=true'
A little more details.
When the server is restarted, the first client connection and selection from this table of course started garbage collection (delete record versions).
But for a 24/7 service this is certainly not a good solution.
The question is, how to correctly and safetly close the transaction that occurs when a ThreadSafeConnection is created?
Thanks!
Addon.
Maybe this is more related to Zeos Database Objects transaction handling than ORM.
Please give your advice.
Thanks!
Unfortunately, both
FDBConnectionParams.ZeosURL.Properties.Add('hard_commit=true')
and
FDBConnectionParams.ZeosURL.Properties.Add('hard_commit=false')
same result
Hi.
Faced such a case.
There is a table with 50 records.
Each record is updated every minute from FHttpClient: TSQLHttpClient, 24/7.
As a result, a lot of record version accumulated for this table, since the ID of the oldest active transaction is equal to the ID of the first update transaction for this table at server start.
Maybe someone knows how to handle this?
Thanks!
Looks like it's already fixed.
Please check https://synopse.info/fossil/info/530adf5475e9af5
ab,
Is it by design, that in mORMotDB.TSQLRestStorageExternal.EngineUpdateField/EngineUpdateFieldIncrement no update of TRecordVersion fields?
Thanks.
Pages: 1