Add a dump|load test, fix multiple bugs in both new&old dump/load utils

Details: - New store dump/write-meta didn't use actual metadata parameters from the header - New store write-meta calculated small entry sizes incorrectly - Old store write-journal imported entries with checksums incorrectly - Old store write-meta didn't import block_csums at all (it was using a wrong json key)
2026-06-17 01:19:12 +03:00
parent c491db699c
commit 430d3cfb6f
5 changed files with 145 additions and 17 deletions
@@ -306,6 +306,78 @@ jobs:
          echo ""
        done

+  test_dump_load:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_dump_load.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_dump_load_32k:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: TEST_NAME=32k OSD_ARGS="--data_csum_type crc32c --csum_block_size 32k" OFFSET_ARGS="$OSD_ARGS" /root/vitastor/tests/test_dump_load.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_old_dump_load:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: OLD=1 /root/vitastor/tests/test_dump_load.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_dump_load_old_32k:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: TEST_NAME=old_32k OLD=1 OSD_ARGS="--data_csum_type crc32c --csum_block_size 32k" OFFSET_ARGS="$OSD_ARGS" /root/vitastor/tests/test_dump_load.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
  test_old_interrupted_rebalance:
    runs-on: ubuntu-latest
    needs: build
@@ -477,11 +477,15 @@ int disk_tool_t::write_json_journal(json11::Json entries)
        uint16_t type = t_it->second;
        if (type == JE_START)
            continue;
+        uint32_t offset = (uint32_t)rec["offset"].uint64_value();
+        uint32_t len = (uint32_t)rec["len"].uint64_value();
+        uint32_t data_csum_blocks = !dsk.data_csum_type || !len ? 0 :
+            (((offset + len - 1)/dsk.csum_block_size - offset/dsk.csum_block_size + 1));
+        uint32_t data_csum_size = data_csum_blocks*(dsk.data_csum_type & 0xFF);
        uint32_t entry_size = (type == JE_START
            ? sizeof(journal_entry_start)
            : (type == JE_SMALL_WRITE || type == JE_SMALL_WRITE_INSTANT
-                ? sizeof(journal_entry_small_write) + dsk.clean_entry_bitmap_size +
-                    (dsk.data_csum_type ? rec["len"].uint64_value()/dsk.csum_block_size*(dsk.data_csum_type & 0xFF) : 0)
+                ? sizeof(journal_entry_small_write) + dsk.clean_entry_bitmap_size + data_csum_size
                : (type == JE_BIG_WRITE || type == JE_BIG_WRITE_INSTANT
                    ? sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size +
                        (dsk.data_csum_type ? rec["len"].uint64_value()/dsk.csum_block_size*(dsk.data_csum_type & 0xFF) : 0)
@@ -523,15 +527,12 @@ int disk_tool_t::write_json_journal(json11::Json entries)
                    .stripe = sscanf_json(NULL, rec["stripe"]),
                },
                .version = rec["ver"].uint64_value(),
-                .offset = (uint32_t)rec["offset"].uint64_value(),
-                .len = (uint32_t)rec["len"].uint64_value(),
+                .offset = offset,
+                .len = len,
                .data_offset = (uint64_t)(new_journal_data-new_journal_buf),
                .crc32_data = !dsk.data_csum_type ? 0 : (uint32_t)sscanf_json("%x", rec["data_crc32"]),
            };
-            uint32_t data_csum_blocks = !dsk.data_csum_type ? 0 :
-                (((ne->small_write.offset+ne->small_write.len)/dsk.csum_block_size - ne->small_write.len/dsk.csum_block_size));
-            uint32_t data_csum_size = data_csum_blocks*(dsk.data_csum_type & 0xFF);
-            fromhexstr(rec["bitmap"].string_value(), dsk.clean_entry_bitmap_size, ((uint8_t*)ne) + sizeof(journal_entry_small_write) + data_csum_size);
+            fromhexstr(rec["bitmap"].string_value(), dsk.clean_entry_bitmap_size, ((uint8_t*)ne) + sizeof(journal_entry_small_write));
            fromhexstr(rec["data"].string_value(), ne->small_write.len, new_journal_data);
            if (ne->small_write.len > 0 && !rec["data"].is_string())
            {
@@ -545,7 +546,7 @@ int disk_tool_t::write_json_journal(json11::Json entries)
                    ne->small_write.crc32_data = crc32c(0, new_journal_data, ne->small_write.len);
                else if (dsk.data_csum_type == BLOCKSTORE_CSUM_CRC32C)
                {
-                    uint32_t *block_csums = (uint32_t*)(((uint8_t*)ne) + sizeof(journal_entry_small_write));
+                    uint32_t *block_csums = (uint32_t*)(((uint8_t*)ne) + sizeof(journal_entry_small_write) + dsk.clean_entry_bitmap_size);
                    for (uint32_t i = 0; i < data_csum_blocks; i++)
                    {
                        uint32_t block_begin = (ne->small_write.offset/dsk.csum_block_size + i) * dsk.csum_block_size;
@@ -574,12 +575,9 @@ int disk_tool_t::write_json_journal(json11::Json entries)
                .len = (uint32_t)rec["len"].uint64_value(),
                .location = sscanf_json(NULL, rec["loc"]),
            };
-            uint32_t data_csum_blocks = !dsk.data_csum_type ? 0 :
-                (((ne->small_write.offset+ne->small_write.len)/dsk.csum_block_size - ne->small_write.len/dsk.csum_block_size));
-            uint32_t data_csum_size = data_csum_blocks*(dsk.data_csum_type & 0xFF);
-            fromhexstr(rec["bitmap"].string_value(), dsk.clean_entry_bitmap_size, ((uint8_t*)ne) + sizeof(journal_entry_big_write) + data_csum_size);
+            fromhexstr(rec["bitmap"].string_value(), dsk.clean_entry_bitmap_size, ((uint8_t*)ne) + sizeof(journal_entry_big_write));
            if (dsk.data_csum_type)
-                fromhexstr(rec["block_csums"].string_value(), data_csum_size, ((uint8_t*)ne) + sizeof(journal_entry_big_write));
+                fromhexstr(rec["block_csums"].string_value(), data_csum_size, ((uint8_t*)ne) + sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size);
        }
        else if (type == JE_STABLE || type == JE_ROLLBACK || type == JE_DELETE)
        {
@@ -95,6 +95,12 @@ close_error:
                journal_pos += read_len;
            }
        }
+        dsk.meta_format = hdr->version;
+        dsk.data_block_size = hdr->data_block_size;
+        dsk.csum_block_size = hdr->csum_block_size;
+        dsk.data_csum_type = hdr->data_csum_type;
+        dsk.bitmap_granularity = hdr->bitmap_granularity;
+        dsk.clean_entry_bitmap_size = (hdr->data_block_size / hdr->bitmap_granularity + 7) / 8;
        blockstore_heap_t *heap = new blockstore_heap_t(&dsk, buffer_area, log_level);
        // Load heap and just iterate it in memory
        hdr_fn(hdr);
@@ -575,7 +581,7 @@ int disk_tool_t::write_json_meta(json11::Json meta)
        {
            if (new_data_csum_size)
            {
-                fromhexstr(e["data_csum"].string_value(), new_data_csum_size,
+                fromhexstr(e["block_csums"].string_value(), new_data_csum_size,
                    ((uint8_t*)new_entry) + sizeof(clean_disk_entry) + 2*new_clean_entry_bitmap_size);
            }
            uint32_t *new_entry_csum = (uint32_t*)(((uint8_t*)new_entry) + new_clean_entry_size - 4);
@@ -616,6 +622,12 @@ int disk_tool_t::write_json_heap(json11::Json meta, json11::Json journal)
    new_data_csum_size = (new_meta_hdr->csum_block_size
        ? ((new_meta_hdr->data_block_size+new_meta_hdr->csum_block_size-1)/new_meta_hdr->csum_block_size*(new_meta_hdr->data_csum_type & 0xFF))
        : 0);
+    dsk.meta_format = new_meta_hdr->version;
+    dsk.data_block_size = new_meta_hdr->data_block_size;
+    dsk.csum_block_size = new_meta_hdr->csum_block_size;
+    dsk.data_csum_type = new_meta_hdr->data_csum_type;
+    dsk.bitmap_granularity = new_meta_hdr->bitmap_granularity;
+    dsk.clean_entry_bitmap_size = (new_meta_hdr->data_block_size / new_meta_hdr->bitmap_granularity + 7) / 8;
    new_journal_buf = NULL;
    if (new_journal_len)
    {
@@ -696,7 +708,7 @@ close_err0:
                wr->entry_type = wr_type | (write_entry["stable"].bool_value() ? BS_HEAP_STABLE : 0);
                wr->lsn = write_entry["lsn"].uint64_value();
                wr->version = write_entry["version"].uint64_value();
-                wr->size = wr->get_size(&heap);
+                wr->size = wr_size;
                if (wr_type == BS_HEAP_SMALL_WRITE || wr_type == BS_HEAP_INTENT_WRITE)
                {
                    wr->small().offset = wr_offset;
@@ -736,6 +748,7 @@ close_err0:
                    bi.offset = wr_offset;
                    bi.len = wr_len;
                }
+                wr->size = wr->get_size(&heap);
                if (write_entry["bitmap"].is_string() && wr->get_int_bitmap(&heap))
                {
                    fromhexstr(write_entry["bitmap"].string_value(), new_clean_entry_bitmap_size, wr->get_int_bitmap(&heap));
@@ -794,7 +807,7 @@ close_err:
            fromhexstr(meta_entry["bitmap"].string_value(), new_clean_entry_bitmap_size, wr->get_int_bitmap(&heap));
            fromhexstr(meta_entry["ext_bitmap"].string_value(), new_clean_entry_bitmap_size, wr->get_ext_bitmap(&heap));
            if (new_data_csum_size)
-                fromhexstr(meta_entry["data_csum"].string_value(), new_data_csum_size, wr->get_checksums(&heap));
+                fromhexstr(meta_entry["block_csums"].string_value(), new_data_csum_size, wr->get_checksums(&heap));
            wr->crc32c = wr->calc_crc32c();
            assert((uint8_t*)wr + wr->size == new_meta_buf + meta_offset + used_space);
            auto j_it = journal_by_object.find(oid);
@@ -24,6 +24,11 @@ IMMEDIATE_COMMIT=1 ./test_interrupted_rebalance.sh
 SCHEME=ec ./test_interrupted_rebalance.sh
 SCHEME=ec IMMEDIATE_COMMIT=1 ./test_interrupted_rebalance.sh

+./test_dump_load.sh
+TEST_NAME=32k OSD_ARGS="--data_csum_type crc32c --csum_block_size 32k" OFFSET_ARGS="$OSD_ARGS" ./test_dump_load.sh
+OLD=1 ./test_dump_load.sh
+TEST_NAME=old_32k OLD=1 OSD_ARGS="--data_csum_type crc32c --csum_block_size 32k" OFFSET_ARGS="$OSD_ARGS" ./test_dump_load.sh
+
 OLD=1 ./test_interrupted_rebalance.sh
 OLD=1 IMMEDIATE_COMMIT=1 ./test_interrupted_rebalance.sh
 OLD=1 SCHEME=ec ./test_interrupted_rebalance.sh
@@ -0,0 +1,40 @@
+#!/bin/bash -ex
+
+PG_COUNT=${PG_COUNT:-32}
+
+. `dirname $0`/run_3osds.sh
+
+$VITASTOR_FIO -bs=4M -direct=1 -iodepth=4 \
+    -rw=write -end_fsync=1 -pool=1 -inode=1 -size=256M -runtime=10
+
+$VITASTOR_FIO -bs=4k -direct=1 -iodepth=32 \
+    -rw=randwrite -end_fsync=1 -pool=1 -inode=1 -size=256M -runtime=10 -number_ios=1024
+
+for i in $(seq 1 $OSD_COUNT); do
+    pid=OSD${i}_PID
+    pid=${!pid}
+    kill -9 $pid
+done
+
+offsets=$(build/src/disk_tool/vitastor-disk simple-offsets --format json ./testdata/bin/test_osd1.bin $OFFSET_ARGS)
+opts=$(build/src/disk_tool/vitastor-disk simple-offsets --format options ./testdata/bin/test_osd1.bin $OFFSET_ARGS)
+meta_format=$(echo $offsets | jq -r .meta_format)
+journal_offset=$(echo $offsets | jq -r .journal_offset)
+meta_offset=$(echo $offsets | jq -r .meta_offset)
+data_offset=$(echo $offsets | jq -r .data_offset)
+if [[ "$meta_format" = "3" ]]; then
+    build/src/disk_tool/vitastor-disk dump-meta --io cached $opts | jq '. + { "entries": .entries | sort_by(.stripe) }' >./testdata/meta.json
+    build/src/disk_tool/vitastor-disk write-meta --io cached $opts <./testdata/meta.json
+    build/src/disk_tool/vitastor-disk dump-meta --io cached $opts | jq '. + { "entries": .entries | sort_by(.stripe) }' >./testdata/meta2.json
+else
+    build/src/disk_tool/vitastor-disk dump-journal --io cached --json ./testdata/bin/test_osd1.bin 4096 $journal_offset $((meta_offset-journal_offset)) | jq 'map(del(.crc32, .crc32_prev)) | map(if .type == "small_write" or .type == "small_write_instant" then del(.loc) elif .type == "start" then del(.start) else . end)' >./testdata/journal.json
+    build/src/disk_tool/vitastor-disk write-journal --io cached --json ./testdata/bin/test_osd1.bin 4096 4 $journal_offset $((meta_offset-journal_offset)) <./testdata/journal.json
+    build/src/disk_tool/vitastor-disk dump-journal --io cached --json ./testdata/bin/test_osd1.bin 4096 $journal_offset $((meta_offset-journal_offset)) | jq 'map(del(.crc32, .crc32_prev)) | map(if .type == "small_write" or .type == "small_write_instant" then del(.loc) elif .type == "start" then del(.start) else . end)' >./testdata/journal2.json
+    diff ./testdata/journal.json ./testdata/journal2.json
+    build/src/disk_tool/vitastor-disk dump-meta --io cached ./testdata/bin/test_osd1.bin 4096 $meta_offset $((data_offset-meta_offset)) >./testdata/meta.json
+    build/src/disk_tool/vitastor-disk write-meta --io cached $opts <./testdata/meta.json
+    build/src/disk_tool/vitastor-disk dump-meta --io cached ./testdata/bin/test_osd1.bin 4096 $meta_offset $((data_offset-meta_offset)) >./testdata/meta2.json
+fi
+diff ./testdata/meta.json ./testdata/meta2.json
+
+format_green OK