Volumes related refactorings #12670

excitoon · 2020-07-22T17:22:37Z

I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en

Changelog category (leave one):

Not for changelog (changelog entry is not required)

Detailed description / Documentation draft:

Generic volume

By adding documentation, you'll allow users to try your new feature immediately, not when someone else will have time to document it later. Documentation is necessary for all features that affect user experience in any way. You can add brief documentation draft above, or add documentation right into your patch as Markdown files in docs folder.

If you are doing this for the first time, it's recommended to read the lightweight Contributing to ClickHouse Documentation guide first.

1. Implemented MultiDiskReservation to maintain multiple reservations under one 2. Implemented VolumeRAID1 as child of VolumeJBOD (since RAID1 is quite similar) 3. Implemented createVolumeFromConfig

…Config

1. SingleDiskVolume for temporary volumes 2. Generic VolumePtr in StoragePolicies 3. Removed max_data_part_size in system.storage_policies, added volume_type

alesapin

At least, we need to disable RAID1 for users until it will be ready.

alesapin · 2020-07-27T11:02:42Z

src/Disks/VolumeRAID1.h

+};
+
+using VolumeRAID1Ptr = std::shared_ptr<VolumeRAID1>;
+using VolumesRAID1Ptr = std::vector<VolumeRAID1Ptr>;


Not a pointer: VolumesRAID1.

I would remove that line completely.

alesapin · 2020-07-27T11:03:02Z

src/Disks/IVolume.cpp

+
+void MultiDiskReservation::update(UInt64 new_size)
+{
+    for (auto &r: reservations)


alesapin · 2020-07-27T11:03:08Z

src/Disks/IVolume.cpp

+        throw Exception("At least one reservation must be provided to MultiDiskReservation", ErrorCodes::NO_RESERVATIONS_PROVIDED);
+    }
+
+    for (auto &r: reservations)


alesapin · 2020-07-27T11:03:12Z

src/Disks/IVolume.cpp

    return res;
 }

+MultiDiskReservation::MultiDiskReservation(Reservations &reservations_, UInt64 size_)


alesapin · 2020-07-27T11:05:32Z

src/Interpreters/Aggregator.cpp

 void Aggregator::writeToTemporaryFile(AggregatedDataVariants & data_variants)
 {
-    return writeToTemporaryFile(data_variants, params.tmp_volume->getNextDisk()->getPath());
+    return writeToTemporaryFile(data_variants, params.tmp_volume->getDisk()->getPath());


Why we have different logic here and https://github.com/ClickHouse/ClickHouse/pull/12670/files#diff-bafb5527e76244c7f83100af0572d862R771 here?

Looks like a possible bug. @excitoon let's try copy logic from the link and check integration tests (I remember this was part of a test fix)

Good catch!

There is no such thing as getNextDisk() in IVolume. What shall we do? @NanoBjorn

I did as you advised, but that looks ugly. Can we use some of polymorphism here?

I did as you advised, but that looks ugly. Can we use some of polymorphism here?

I thought about it for a long time, this seems the best way, despite the fact it is a bit ugly. So leave it this way:)

alesapin · 2020-07-27T11:07:25Z

src/Storages/StorageDistributed.cpp

-    for (const DiskPtr & disk : volume->getDisks())
+    for (const DiskPtr & disk : storage_policy->getDisks())
    {
+        disk->reserve(0);


Why we need it now?

I don't remember making these changes. @excitoon why are you moving back from volume as member class to storage_policy?

Would it be better if I use reserve() from volume? @NanoBjorn
This is 1b3f5c9

I still don't understand why we needed to move back from volume as member to storage_policy, but maybe it is okay if it works. Why you need reserve here at all? It was fine without it, or not?

alesapin · 2020-07-27T11:10:42Z

src/Storages/System/StorageSystemStoragePolicies.cpp

             {"volume_name", std::make_shared<DataTypeString>()},
             {"volume_priority", std::make_shared<DataTypeUInt64>()},
             {"disks", std::make_shared<DataTypeArray>(std::make_shared<DataTypeString>())},
-             {"max_data_part_size", std::make_shared<DataTypeUInt64>()},


Why we lost it?

I guess if we want it here then we should move max_data_part_size to IVolume. @alesapin do we want it in IVolume or not?

Yes this is sad. @NanoBjorn

I've taken max_data_part_size back. @alesapin please check.

alesapin · 2020-07-28T10:44:24Z

src/Disks/createVolume.cpp

+        return std::make_shared<VolumeJBOD>(name, config, config_prefix, disk_selector);
+    }
+    String raid_type = config.getString(config_prefix + ".raid_type");
+    if (raid_type == "RAID 1" || raid_type == "RAID-1" || raid_type == "RAID1")


One more time: It shouldn't be possible to create RAID1 in config.xml. Also, we need to add a comment that it doesn't work yet.

Why do we support ambiguity in the name of RAID-1? Let's stick with only one variant.

@excitoon just remove all lines devoted to RAID 1, that was my mistake, we don't need it in this PR

It's been almost 2 years since it was added, and it hadn't been implemented yet.
Maybe it's time to remove it? (#36836)

alexey-milovidov · 2020-07-28T10:58:19Z

Docs check — Found 1 errors

Broken by @blinkov

Merging #12670

azat · 2020-09-16T08:43:46Z

src/Storages/StorageDistributed.cpp


        const auto & sharding_key = engine_args.size() >= 4 ? engine_args[3] : nullptr;
-        const auto & storage_policy = engine_args.size() >= 5 ? engine_args[4]->as<ASTLiteral &>().value.safeGet<String>() : "";
+        const auto & storage_policy = engine_args.size() >= 5 ? engine_args[4]->as<ASTLiteral &>().value.safeGet<String>() : "default";


Actually this wasn't the "redundant code" (f5af645), now it uses default storage policy even if the 5-th argument is not passed, while previous behavior creates separate volume with just path from config.xml

excitoon mentioned this pull request Jul 22, 2020

Volumes related refactorings #11883

Closed

NanoBjorn and others added 16 commits July 26, 2020 21:27

MultiDiskReservation, VolumeRAID1 and createVolumeFromConfig

3585b24

1. Implemented MultiDiskReservation to maintain multiple reservations under one 2. Implemented VolumeRAID1 as child of VolumeJBOD (since RAID1 is quite similar) 3. Implemented createVolumeFromConfig

Style fixes and added UNKNOWN_RAID_TYPE exception to createVolumeFrom…

db56820

…Config

Small clang-10 warning fixes + ya.make

96c7cc8

Generic volume is coming...

7f5b6fb

1. SingleDiskVolume for temporary volumes 2. Generic VolumePtr in StoragePolicies 3. Removed max_data_part_size in system.storage_policies, added volume_type

Small fix of VolumeRAID1Ptr

adef249

Braces style fix

2c39584

removed default clause in switch for VolumeType enum

e4699e2

removed unused error code

f4080ed

Small fixes for tests

99d5255

Fixed integration test_tmp_policy

aac9795

Fixed integration test_multiple_disks

ee07c74

Minor fix in StorageDistributed.

8039d45

Review fixes.

ee21fde

Test fix (removed redundant code).

f5af645

Real fix of test.

1b3f5c9

Test fix.

a9e0a5d

alesapin requested changes Jul 27, 2020

View reviewed changes

alexey-milovidov assigned alesapin Jul 27, 2020

excitoon added 3 commits July 27, 2020 18:59

Review fixes.

6d22069

Minor thing.

83556a5

Empty commit to re-run checks.

98fca0a

alesapin requested changes Jul 28, 2020

View reviewed changes

excitoon added 2 commits July 28, 2020 19:45

Minor fixes.

faedb04

Returned max_data_part_size to system.storage_policies.

bcc926f

excitoon requested a review from alesapin July 30, 2020 01:19

alesapin mentioned this pull request Jul 30, 2020

Merging #12670 #13104

Merged

alesapin added a commit that referenced this pull request Jul 31, 2020

Merge pull request #13104 from ClickHouse/volumes-refactorings

2eb01fd

Merging #12670

alesapin merged commit 724b7be into ClickHouse:master Jul 31, 2020

filimonov added the altinity label Aug 4, 2020

azat mentioned this pull request Sep 16, 2020

Don't throw exception if Distributed storage has multi-volume storage policy configuration #14839

Merged

azat reviewed Sep 16, 2020

View reviewed changes

Volumes related refactorings #12670

Volumes related refactorings #12670

Uh oh!

Conversation

excitoon commented Jul 22, 2020

Uh oh!

alesapin left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

NanoBjorn Jul 27, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alexey-milovidov commented Jul 28, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

NanoBjorn Jul 27, 2020 •

edited

Loading